linux/net/netlink/genetlink.c

1816 lines
41 KiB
C
Raw Normal View History

License cleanup: add SPDX GPL-2.0 license identifier to files with no license Many source files in the tree are missing licensing information, which makes it harder for compliance tools to determine the correct license. By default all files without license information are under the default license of the kernel, which is GPL version 2. Update the files which contain no license information with the 'GPL-2.0' SPDX license identifier. The SPDX identifier is a legally binding shorthand, which can be used instead of the full boiler plate text. This patch is based on work done by Thomas Gleixner and Kate Stewart and Philippe Ombredanne. How this work was done: Patches were generated and checked against linux-4.14-rc6 for a subset of the use cases: - file had no licensing information it it. - file was a */uapi/* one with no licensing information in it, - file was a */uapi/* one with existing licensing information, Further patches will be generated in subsequent months to fix up cases where non-standard license headers were used, and references to license had to be inferred by heuristics based on keywords. The analysis to determine which SPDX License Identifier to be applied to a file was done in a spreadsheet of side by side results from of the output of two independent scanners (ScanCode & Windriver) producing SPDX tag:value files created by Philippe Ombredanne. Philippe prepared the base worksheet, and did an initial spot review of a few 1000 files. The 4.13 kernel was the starting point of the analysis with 60,537 files assessed. Kate Stewart did a file by file comparison of the scanner results in the spreadsheet to determine which SPDX license identifier(s) to be applied to the file. She confirmed any determination that was not immediately clear with lawyers working with the Linux Foundation. Criteria used to select files for SPDX license identifier tagging was: - Files considered eligible had to be source code files. - Make and config files were included as candidates if they contained >5 lines of source - File already had some variant of a license header in it (even if <5 lines). All documentation files were explicitly excluded. The following heuristics were used to determine which SPDX license identifiers to apply. - when both scanners couldn't find any license traces, file was considered to have no license information in it, and the top level COPYING file license applied. For non */uapi/* files that summary was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 11139 and resulted in the first patch in this series. If that file was a */uapi/* path one, it was "GPL-2.0 WITH Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 WITH Linux-syscall-note 930 and resulted in the second patch in this series. - if a file had some form of licensing information in it, and was one of the */uapi/* ones, it was denoted with the Linux-syscall-note if any GPL family license was found in the file or had no licensing in it (per prior point). Results summary: SPDX license identifier # files ---------------------------------------------------|------ GPL-2.0 WITH Linux-syscall-note 270 GPL-2.0+ WITH Linux-syscall-note 169 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17 LGPL-2.1+ WITH Linux-syscall-note 15 GPL-1.0+ WITH Linux-syscall-note 14 ((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5 LGPL-2.0+ WITH Linux-syscall-note 4 LGPL-2.1 WITH Linux-syscall-note 3 ((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3 ((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1 and that resulted in the third patch in this series. - when the two scanners agreed on the detected license(s), that became the concluded license(s). - when there was disagreement between the two scanners (one detected a license but the other didn't, or they both detected different licenses) a manual inspection of the file occurred. - In most cases a manual inspection of the information in the file resulted in a clear resolution of the license that should apply (and which scanner probably needed to revisit its heuristics). - When it was not immediately clear, the license identifier was confirmed with lawyers working with the Linux Foundation. - If there was any question as to the appropriate license identifier, the file was flagged for further research and to be revisited later in time. In total, over 70 hours of logged manual review was done on the spreadsheet to determine the SPDX license identifiers to apply to the source files by Kate, Philippe, Thomas and, in some cases, confirmation by lawyers working with the Linux Foundation. Kate also obtained a third independent scan of the 4.13 code base from FOSSology, and compared selected files where the other two scanners disagreed against that SPDX file, to see if there was new insights. The Windriver scanner is based on an older version of FOSSology in part, so they are related. Thomas did random spot checks in about 500 files from the spreadsheets for the uapi headers and agreed with SPDX license identifier in the files he inspected. For the non-uapi files Thomas did random spot checks in about 15000 files. In initial set of patches against 4.14-rc6, 3 files were found to have copy/paste license identifier errors, and have been fixed to reflect the correct identifier. Additionally Philippe spent 10 hours this week doing a detailed manual inspection and review of the 12,461 patched files from the initial patch version early this week with: - a full scancode scan run, collecting the matched texts, detected license ids and scores - reviewing anything where there was a license detected (about 500+ files) to ensure that the applied SPDX license was correct - reviewing anything where there was no detection but the patch license was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied SPDX license was correct This produced a worksheet with 20 files needing minor correction. This worksheet was then exported into 3 different .csv files for the different types of files to be modified. These .csv files were then reviewed by Greg. Thomas wrote a script to parse the csv files and add the proper SPDX tag to the file, in the format that the file expected. This script was further refined by Greg based on the output to detect more types of files automatically and to distinguish between header and source .c files (which need different comment types.) Finally Greg ran the script using the .csv files to generate the patches. Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-11-01 17:07:57 +03:00
// SPDX-License-Identifier: GPL-2.0
/*
* NETLINK Generic Netlink Family
*
* Authors: Jamal Hadi Salim
* Thomas Graf <tgraf@suug.ch>
* Johannes Berg <johannes@sipsolutions.net>
*/
#include <linux/module.h>
#include <linux/kernel.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 11:04:11 +03:00
#include <linux/slab.h>
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/string.h>
#include <linux/skbuff.h>
#include <linux/mutex.h>
#include <linux/bitmap.h>
#include <linux/rwsem.h>
#include <linux/idr.h>
#include <net/sock.h>
#include <net/genetlink.h>
static DEFINE_MUTEX(genl_mutex); /* serialization of message processing */
static DECLARE_RWSEM(cb_lock);
genetlink: synchronize socket closing and family removal In addition to the problem Jeff Layton reported, I looked at the code and reproduced the same warning by subscribing and removing the genl family with a socket still open. This is a fairly tricky race which originates in the fact that generic netlink allows the family to go away while sockets are still open - unlike regular netlink which has a module refcount for every open socket so in general this cannot be triggered. Trying to resolve this issue by the obvious locking isn't possible as it will result in deadlocks between unregistration and group unbind notification (which incidentally lockdep doesn't find due to the home grown locking in the netlink table.) To really resolve this, introduce a "closing socket" reference counter (for generic netlink only, as it's the only affected family) in the core netlink code and use that in generic netlink to wait for all the sockets that are being closed at the same time as a generic netlink family is removed. This fixes the race that when a socket is closed, it will should call the unbind, but if the family is removed at the same time the unbind will not find it, leading to the warning. The real problem though is that in this case the unbind could actually find a new family that is registered to have a multicast group with the same ID, and call its mcast_unbind() leading to confusing. Also remove the warning since it would still trigger, but is now no longer a problem. This also moves the code in af_netlink.c to before unreferencing the module to avoid having the same problem in the normal non-genl case. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 13:37:14 +03:00
atomic_t genl_sk_destructing_cnt = ATOMIC_INIT(0);
DECLARE_WAIT_QUEUE_HEAD(genl_sk_destructing_waitq);
void genl_lock(void)
{
mutex_lock(&genl_mutex);
}
EXPORT_SYMBOL(genl_lock);
void genl_unlock(void)
{
mutex_unlock(&genl_mutex);
}
EXPORT_SYMBOL(genl_unlock);
static void genl_lock_all(void)
{
down_write(&cb_lock);
genl_lock();
}
static void genl_unlock_all(void)
{
genl_unlock();
up_write(&cb_lock);
}
static DEFINE_IDR(genl_fam_idr);
/*
* Bitmap of multicast groups that are currently in use.
*
* To avoid an allocation at boot of just one unsigned long,
* declare it global instead.
* Bit 0 is marked as already used since group 0 is invalid.
* Bit 1 is marked as already used since the drop-monitor code
* abuses the API and thinks it can statically use group 1.
* That group will typically conflict with other groups that
* any proper users use.
* Bit 16 is marked as used since it's used for generic netlink
* and the code no longer marks pre-reserved IDs as used.
* Bit 17 is marked as already used since the VFS quota code
* also abused this API and relied on family == group ID, we
* cater to that by giving it a static family and group ID.
* Bit 18 is marked as already used since the PMCRAID driver
* did the same thing as the VFS quota code (maybe copied?)
*/
static unsigned long mc_group_start = 0x3 | BIT(GENL_ID_CTRL) |
BIT(GENL_ID_VFS_DQUOT) |
BIT(GENL_ID_PMCRAID);
static unsigned long *mc_groups = &mc_group_start;
static unsigned long mc_groups_longs = 1;
/* We need the last attribute with non-zero ID therefore a 2-entry array */
static struct nla_policy genl_policy_reject_all[] = {
{ .type = NLA_REJECT },
{ .type = NLA_REJECT },
};
static int genl_ctrl_event(int event, const struct genl_family *family,
const struct genl_multicast_group *grp,
int grp_id);
static void
genl_op_fill_in_reject_policy(const struct genl_family *family,
struct genl_ops *op)
{
BUILD_BUG_ON(ARRAY_SIZE(genl_policy_reject_all) - 1 != 1);
if (op->policy || op->cmd < family->resv_start_op)
return;
op->policy = genl_policy_reject_all;
op->maxattr = 1;
}
static void
genl_op_fill_in_reject_policy_split(const struct genl_family *family,
struct genl_split_ops *op)
{
if (op->policy)
return;
op->policy = genl_policy_reject_all;
op->maxattr = 1;
}
static const struct genl_family *genl_family_find_byid(unsigned int id)
{
return idr_find(&genl_fam_idr, id);
}
static const struct genl_family *genl_family_find_byname(char *name)
{
const struct genl_family *family;
unsigned int id;
idr_for_each_entry(&genl_fam_idr, family, id)
if (strcmp(family->name, name) == 0)
return family;
return NULL;
}
struct genl_op_iter {
const struct genl_family *family;
struct genl_split_ops doit;
struct genl_split_ops dumpit;
int cmd_idx;
int entry_idx;
u32 cmd;
u8 flags;
};
static void genl_op_from_full(const struct genl_family *family,
unsigned int i, struct genl_ops *op)
{
*op = family->ops[i];
if (!op->maxattr)
op->maxattr = family->maxattr;
if (!op->policy)
op->policy = family->policy;
genl_op_fill_in_reject_policy(family, op);
}
static int genl_get_cmd_full(u32 cmd, const struct genl_family *family,
struct genl_ops *op)
{
int i;
for (i = 0; i < family->n_ops; i++)
if (family->ops[i].cmd == cmd) {
genl_op_from_full(family, i, op);
return 0;
}
return -ENOENT;
}
static void genl_op_from_small(const struct genl_family *family,
unsigned int i, struct genl_ops *op)
{
memset(op, 0, sizeof(*op));
op->doit = family->small_ops[i].doit;
op->dumpit = family->small_ops[i].dumpit;
op->cmd = family->small_ops[i].cmd;
op->internal_flags = family->small_ops[i].internal_flags;
op->flags = family->small_ops[i].flags;
op->validate = family->small_ops[i].validate;
op->maxattr = family->maxattr;
op->policy = family->policy;
genl_op_fill_in_reject_policy(family, op);
}
static int genl_get_cmd_small(u32 cmd, const struct genl_family *family,
struct genl_ops *op)
{
int i;
for (i = 0; i < family->n_small_ops; i++)
if (family->small_ops[i].cmd == cmd) {
genl_op_from_small(family, i, op);
return 0;
}
return -ENOENT;
}
static void genl_op_from_split(struct genl_op_iter *iter)
{
const struct genl_family *family = iter->family;
int i, cnt = 0;
i = iter->entry_idx - family->n_ops - family->n_small_ops;
if (family->split_ops[i + cnt].flags & GENL_CMD_CAP_DO) {
iter->doit = family->split_ops[i + cnt];
genl_op_fill_in_reject_policy_split(family, &iter->doit);
cnt++;
} else {
memset(&iter->doit, 0, sizeof(iter->doit));
}
if (i + cnt < family->n_split_ops &&
family->split_ops[i + cnt].flags & GENL_CMD_CAP_DUMP) {
iter->dumpit = family->split_ops[i + cnt];
genl_op_fill_in_reject_policy_split(family, &iter->dumpit);
cnt++;
} else {
memset(&iter->dumpit, 0, sizeof(iter->dumpit));
}
WARN_ON(!cnt);
iter->entry_idx += cnt;
}
static int
genl_get_cmd_split(u32 cmd, u8 flag, const struct genl_family *family,
struct genl_split_ops *op)
{
int i;
for (i = 0; i < family->n_split_ops; i++)
if (family->split_ops[i].cmd == cmd &&
family->split_ops[i].flags & flag) {
*op = family->split_ops[i];
return 0;
}
return -ENOENT;
}
static int
genl_cmd_full_to_split(struct genl_split_ops *op,
const struct genl_family *family,
const struct genl_ops *full, u8 flags)
{
if ((flags & GENL_CMD_CAP_DO && !full->doit) ||
(flags & GENL_CMD_CAP_DUMP && !full->dumpit)) {
memset(op, 0, sizeof(*op));
return -ENOENT;
}
if (flags & GENL_CMD_CAP_DUMP) {
op->start = full->start;
op->dumpit = full->dumpit;
op->done = full->done;
} else {
op->pre_doit = family->pre_doit;
op->doit = full->doit;
op->post_doit = family->post_doit;
}
if (flags & GENL_CMD_CAP_DUMP &&
full->validate & GENL_DONT_VALIDATE_DUMP) {
op->policy = NULL;
op->maxattr = 0;
} else {
op->policy = full->policy;
op->maxattr = full->maxattr;
}
op->cmd = full->cmd;
op->internal_flags = full->internal_flags;
op->flags = full->flags;
op->validate = full->validate;
/* Make sure flags include the GENL_CMD_CAP_DO / GENL_CMD_CAP_DUMP */
op->flags |= flags;
return 0;
}
/* Must make sure that op is initialized to 0 on failure */
static int
genl_get_cmd(u32 cmd, u8 flags, const struct genl_family *family,
struct genl_split_ops *op)
{
struct genl_ops full;
int err;
err = genl_get_cmd_full(cmd, family, &full);
if (err == -ENOENT)
err = genl_get_cmd_small(cmd, family, &full);
/* Found one of legacy forms */
if (err == 0)
return genl_cmd_full_to_split(op, family, &full, flags);
err = genl_get_cmd_split(cmd, flags, family, op);
if (err)
memset(op, 0, sizeof(*op));
return err;
}
/* For policy dumping only, get ops of both do and dump.
* Fail if both are missing, genl_get_cmd() will zero-init in case of failure.
*/
static int
genl_get_cmd_both(u32 cmd, const struct genl_family *family,
struct genl_split_ops *doit, struct genl_split_ops *dumpit)
{
int err1, err2;
err1 = genl_get_cmd(cmd, GENL_CMD_CAP_DO, family, doit);
err2 = genl_get_cmd(cmd, GENL_CMD_CAP_DUMP, family, dumpit);
return err1 && err2 ? -ENOENT : 0;
}
static bool
genl_op_iter_init(const struct genl_family *family, struct genl_op_iter *iter)
{
iter->family = family;
iter->cmd_idx = 0;
iter->entry_idx = 0;
iter->flags = 0;
return iter->family->n_ops +
iter->family->n_small_ops +
iter->family->n_split_ops;
}
static bool genl_op_iter_next(struct genl_op_iter *iter)
{
const struct genl_family *family = iter->family;
bool legacy_op = true;
struct genl_ops op;
if (iter->entry_idx < family->n_ops) {
genl_op_from_full(family, iter->entry_idx, &op);
} else if (iter->entry_idx < family->n_ops + family->n_small_ops) {
genl_op_from_small(family, iter->entry_idx - family->n_ops,
&op);
} else if (iter->entry_idx <
family->n_ops + family->n_small_ops + family->n_split_ops) {
legacy_op = false;
/* updates entry_idx */
genl_op_from_split(iter);
} else {
return false;
}
iter->cmd_idx++;
if (legacy_op) {
iter->entry_idx++;
genl_cmd_full_to_split(&iter->doit, family,
&op, GENL_CMD_CAP_DO);
genl_cmd_full_to_split(&iter->dumpit, family,
&op, GENL_CMD_CAP_DUMP);
}
iter->cmd = iter->doit.cmd | iter->dumpit.cmd;
iter->flags = iter->doit.flags | iter->dumpit.flags;
return true;
}
static void
genl_op_iter_copy(struct genl_op_iter *dst, struct genl_op_iter *src)
{
*dst = *src;
}
static unsigned int genl_op_iter_idx(struct genl_op_iter *iter)
{
return iter->cmd_idx;
}
static int genl_allocate_reserve_groups(int n_groups, int *first_id)
{
unsigned long *new_groups;
int start = 0;
int i;
int id;
bool fits;
do {
if (start == 0)
id = find_first_zero_bit(mc_groups,
mc_groups_longs *
BITS_PER_LONG);
else
id = find_next_zero_bit(mc_groups,
mc_groups_longs * BITS_PER_LONG,
start);
fits = true;
for (i = id;
i < min_t(int, id + n_groups,
mc_groups_longs * BITS_PER_LONG);
i++) {
if (test_bit(i, mc_groups)) {
start = i;
fits = false;
break;
}
}
if (id + n_groups > mc_groups_longs * BITS_PER_LONG) {
unsigned long new_longs = mc_groups_longs +
BITS_TO_LONGS(n_groups);
size_t nlen = new_longs * sizeof(unsigned long);
if (mc_groups == &mc_group_start) {
new_groups = kzalloc(nlen, GFP_KERNEL);
if (!new_groups)
return -ENOMEM;
mc_groups = new_groups;
*mc_groups = mc_group_start;
} else {
new_groups = krealloc(mc_groups, nlen,
GFP_KERNEL);
if (!new_groups)
return -ENOMEM;
mc_groups = new_groups;
for (i = 0; i < BITS_TO_LONGS(n_groups); i++)
mc_groups[mc_groups_longs + i] = 0;
}
mc_groups_longs = new_longs;
}
} while (!fits);
for (i = id; i < id + n_groups; i++)
set_bit(i, mc_groups);
*first_id = id;
return 0;
}
static struct genl_family genl_ctrl;
static int genl_validate_assign_mc_groups(struct genl_family *family)
{
int first_id;
int n_groups = family->n_mcgrps;
int err = 0, i;
bool groups_allocated = false;
if (!n_groups)
return 0;
for (i = 0; i < n_groups; i++) {
const struct genl_multicast_group *grp = &family->mcgrps[i];
if (WARN_ON(grp->name[0] == '\0'))
return -EINVAL;
if (WARN_ON(memchr(grp->name, '\0', GENL_NAMSIZ) == NULL))
return -EINVAL;
}
/* special-case our own group and hacks */
if (family == &genl_ctrl) {
first_id = GENL_ID_CTRL;
BUG_ON(n_groups != 1);
} else if (strcmp(family->name, "NET_DM") == 0) {
first_id = 1;
BUG_ON(n_groups != 1);
} else if (family->id == GENL_ID_VFS_DQUOT) {
first_id = GENL_ID_VFS_DQUOT;
BUG_ON(n_groups != 1);
} else if (family->id == GENL_ID_PMCRAID) {
first_id = GENL_ID_PMCRAID;
BUG_ON(n_groups != 1);
} else {
groups_allocated = true;
err = genl_allocate_reserve_groups(n_groups, &first_id);
if (err)
return err;
}
family->mcgrp_offset = first_id;
/* if still initializing, can't and don't need to realloc bitmaps */
if (!init_net.genl_sock)
return 0;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
if (family->netnsok) {
struct net *net;
netlink_table_grab();
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
rcu_read_lock();
for_each_net_rcu(net) {
err = __netlink_change_ngroups(net->genl_sock,
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
mc_groups_longs * BITS_PER_LONG);
if (err) {
/*
* No need to roll back, can only fail if
* memory allocation fails and then the
* number of _possible_ groups has been
* increased on some sockets which is ok.
*/
break;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
}
}
rcu_read_unlock();
netlink_table_ungrab();
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
} else {
err = netlink_change_ngroups(init_net.genl_sock,
mc_groups_longs * BITS_PER_LONG);
}
if (groups_allocated && err) {
for (i = 0; i < family->n_mcgrps; i++)
clear_bit(family->mcgrp_offset + i, mc_groups);
}
return err;
}
static void genl_unregister_mc_groups(const struct genl_family *family)
{
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
struct net *net;
int i;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
netlink_table_grab();
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
rcu_read_lock();
for_each_net_rcu(net) {
for (i = 0; i < family->n_mcgrps; i++)
__netlink_clear_multicast_users(
net->genl_sock, family->mcgrp_offset + i);
}
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
rcu_read_unlock();
netlink_table_ungrab();
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
for (i = 0; i < family->n_mcgrps; i++) {
int grp_id = family->mcgrp_offset + i;
if (grp_id != 1)
clear_bit(grp_id, mc_groups);
genl_ctrl_event(CTRL_CMD_DELMCAST_GRP, family,
&family->mcgrps[i], grp_id);
}
}
static bool genl_split_op_check(const struct genl_split_ops *op)
{
if (WARN_ON(hweight8(op->flags & (GENL_CMD_CAP_DO |
GENL_CMD_CAP_DUMP)) != 1))
return true;
return false;
}
static int genl_validate_ops(const struct genl_family *family)
{
struct genl_op_iter i, j;
unsigned int s;
if (WARN_ON(family->n_ops && !family->ops) ||
WARN_ON(family->n_small_ops && !family->small_ops) ||
WARN_ON(family->n_split_ops && !family->split_ops))
return -EINVAL;
for (genl_op_iter_init(family, &i); genl_op_iter_next(&i); ) {
if (!(i.flags & (GENL_CMD_CAP_DO | GENL_CMD_CAP_DUMP)))
return -EINVAL;
if (WARN_ON(i.cmd >= family->resv_start_op &&
(i.doit.validate || i.dumpit.validate)))
return -EINVAL;
genl_op_iter_copy(&j, &i);
while (genl_op_iter_next(&j)) {
if (i.cmd == j.cmd)
return -EINVAL;
}
}
if (family->n_split_ops) {
if (genl_split_op_check(&family->split_ops[0]))
return -EINVAL;
}
for (s = 1; s < family->n_split_ops; s++) {
const struct genl_split_ops *a, *b;
a = &family->split_ops[s - 1];
b = &family->split_ops[s];
if (genl_split_op_check(b))
return -EINVAL;
/* Check sort order */
if (a->cmd < b->cmd)
continue;
if (a->internal_flags != b->internal_flags ||
((a->flags ^ b->flags) & ~(GENL_CMD_CAP_DO |
GENL_CMD_CAP_DUMP))) {
WARN_ON(1);
return -EINVAL;
}
if ((a->flags & GENL_CMD_CAP_DO) &&
(b->flags & GENL_CMD_CAP_DUMP))
continue;
WARN_ON(1);
return -EINVAL;
}
return 0;
}
/**
* genl_register_family - register a generic netlink family
* @family: generic netlink family
*
* Registers the specified family after validating it first. Only one
* family may be registered with the same family name or identifier.
*
* The family's ops, multicast groups and module pointer must already
* be assigned.
*
* Return 0 on success or a negative error code.
*/
int genl_register_family(struct genl_family *family)
{
int err, i;
int start = GENL_START_ALLOC, end = GENL_MAX_ID;
err = genl_validate_ops(family);
if (err)
return err;
genl_lock_all();
if (genl_family_find_byname(family->name)) {
err = -EEXIST;
goto errout_locked;
}
/*
* Sadly, a few cases need to be special-cased
* due to them having previously abused the API
* and having used their family ID also as their
* multicast group ID, so we use reserved IDs
* for both to be sure we can do that mapping.
*/
if (family == &genl_ctrl) {
/* and this needs to be special for initial family lookups */
start = end = GENL_ID_CTRL;
} else if (strcmp(family->name, "pmcraid") == 0) {
start = end = GENL_ID_PMCRAID;
} else if (strcmp(family->name, "VFS_DQUOT") == 0) {
start = end = GENL_ID_VFS_DQUOT;
}
family->id = idr_alloc_cyclic(&genl_fam_idr, family,
start, end + 1, GFP_KERNEL);
if (family->id < 0) {
err = family->id;
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
goto errout_locked;
}
err = genl_validate_assign_mc_groups(family);
if (err)
goto errout_remove;
genl_unlock_all();
/* send all events */
genl_ctrl_event(CTRL_CMD_NEWFAMILY, family, NULL, 0);
for (i = 0; i < family->n_mcgrps; i++)
genl_ctrl_event(CTRL_CMD_NEWMCAST_GRP, family,
&family->mcgrps[i], family->mcgrp_offset + i);
return 0;
errout_remove:
idr_remove(&genl_fam_idr, family->id);
errout_locked:
genl_unlock_all();
return err;
}
EXPORT_SYMBOL(genl_register_family);
/**
* genl_unregister_family - unregister generic netlink family
* @family: generic netlink family
*
* Unregisters the specified family.
*
* Returns 0 on success or a negative error code.
*/
int genl_unregister_family(const struct genl_family *family)
{
genl_lock_all();
genetlink: Fix generic netlink family unregister This patch fixes a typo in unregister operation. Following crash is fixed by this patch. It can be easily reproduced by repeating modprobe and rmmod module that uses genetlink. [ 261.446686] BUG: unable to handle kernel paging request at ffffffffa0264088 [ 261.448921] IP: [<ffffffff813cb70e>] strcmp+0xe/0x30 [ 261.450494] PGD 1c09067 [ 261.451266] PUD 1c0a063 [ 261.452091] PMD 8068d5067 [ 261.452525] PTE 0 [ 261.453164] [ 261.453618] Oops: 0000 [#1] SMP [ 261.454577] Modules linked in: openvswitch(+) ... [ 261.480753] RIP: 0010:[<ffffffff813cb70e>] [<ffffffff813cb70e>] strcmp+0xe/0x30 [ 261.483069] RSP: 0018:ffffc90003c0bc28 EFLAGS: 00010282 [ 261.510145] Call Trace: [ 261.510896] [<ffffffff816f10ca>] genl_family_find_byname+0x5a/0x70 [ 261.512819] [<ffffffff816f2319>] genl_register_family+0xb9/0x630 [ 261.514805] [<ffffffffa02840bc>] dp_init+0xbc/0x120 [openvswitch] [ 261.518268] [<ffffffff8100217d>] do_one_initcall+0x3d/0x160 [ 261.525041] [<ffffffff811808a9>] do_init_module+0x60/0x1f1 [ 261.526754] [<ffffffff8110687f>] load_module+0x22af/0x2860 [ 261.530144] [<ffffffff81107026>] SYSC_finit_module+0x96/0xd0 [ 261.531901] [<ffffffff8110707e>] SyS_finit_module+0xe/0x10 [ 261.533605] [<ffffffff8100391e>] do_syscall_64+0x6e/0x180 [ 261.535284] [<ffffffff817c2faf>] entry_SYSCALL64_slow_path+0x25/0x25 [ 261.546512] RIP [<ffffffff813cb70e>] strcmp+0xe/0x30 [ 261.550198] ---[ end trace 76505a814dd68770 ]--- Fixes: 2ae0f17df1c ("genetlink: use idr to track families"). Reported-by: Jarno Rajahalme <jarno@ovn.org> CC: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Pravin B Shelar <pshelar@ovn.org> Reviewed-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-10-29 02:01:41 +03:00
if (!genl_family_find_byid(family->id)) {
genl_unlock_all();
return -ENOENT;
}
genl_unregister_mc_groups(family);
genetlink: synchronize socket closing and family removal In addition to the problem Jeff Layton reported, I looked at the code and reproduced the same warning by subscribing and removing the genl family with a socket still open. This is a fairly tricky race which originates in the fact that generic netlink allows the family to go away while sockets are still open - unlike regular netlink which has a module refcount for every open socket so in general this cannot be triggered. Trying to resolve this issue by the obvious locking isn't possible as it will result in deadlocks between unregistration and group unbind notification (which incidentally lockdep doesn't find due to the home grown locking in the netlink table.) To really resolve this, introduce a "closing socket" reference counter (for generic netlink only, as it's the only affected family) in the core netlink code and use that in generic netlink to wait for all the sockets that are being closed at the same time as a generic netlink family is removed. This fixes the race that when a socket is closed, it will should call the unbind, but if the family is removed at the same time the unbind will not find it, leading to the warning. The real problem though is that in this case the unbind could actually find a new family that is registered to have a multicast group with the same ID, and call its mcast_unbind() leading to confusing. Also remove the warning since it would still trigger, but is now no longer a problem. This also moves the code in af_netlink.c to before unreferencing the module to avoid having the same problem in the normal non-genl case. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-16 13:37:14 +03:00
idr_remove(&genl_fam_idr, family->id);
up_write(&cb_lock);
wait_event(genl_sk_destructing_waitq,
atomic_read(&genl_sk_destructing_cnt) == 0);
genl_unlock();
genl_ctrl_event(CTRL_CMD_DELFAMILY, family, NULL, 0);
return 0;
}
EXPORT_SYMBOL(genl_unregister_family);
/**
* genlmsg_put - Add generic netlink header to netlink message
* @skb: socket buffer holding the message
* @portid: netlink portid the message is addressed to
* @seq: sequence number (usually the one of the sender)
* @family: generic netlink family
* @flags: netlink message flags
* @cmd: generic netlink command
*
* Returns pointer to user specific header
*/
void *genlmsg_put(struct sk_buff *skb, u32 portid, u32 seq,
const struct genl_family *family, int flags, u8 cmd)
{
struct nlmsghdr *nlh;
struct genlmsghdr *hdr;
nlh = nlmsg_put(skb, portid, seq, family->id, GENL_HDRLEN +
family->hdrsize, flags);
if (nlh == NULL)
return NULL;
hdr = nlmsg_data(nlh);
hdr->cmd = cmd;
hdr->version = family->version;
hdr->reserved = 0;
return (char *) hdr + GENL_HDRLEN;
}
EXPORT_SYMBOL(genlmsg_put);
static struct genl_dumpit_info *genl_dumpit_info_alloc(void)
{
return kmalloc(sizeof(struct genl_dumpit_info), GFP_KERNEL);
}
static void genl_dumpit_info_free(const struct genl_dumpit_info *info)
{
kfree(info);
}
static struct nlattr **
genl_family_rcv_msg_attrs_parse(const struct genl_family *family,
struct nlmsghdr *nlh,
struct netlink_ext_ack *extack,
const struct genl_split_ops *ops,
int hdrlen,
enum genl_validate_flags no_strict_flag)
{
enum netlink_validation validate = ops->validate & no_strict_flag ?
NL_VALIDATE_LIBERAL :
NL_VALIDATE_STRICT;
struct nlattr **attrbuf;
int err;
if (!ops->maxattr)
return NULL;
attrbuf = kmalloc_array(ops->maxattr + 1,
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
sizeof(struct nlattr *), GFP_KERNEL);
if (!attrbuf)
return ERR_PTR(-ENOMEM);
err = __nlmsg_parse(nlh, hdrlen, attrbuf, ops->maxattr, ops->policy,
validate, extack);
if (err) {
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
kfree(attrbuf);
return ERR_PTR(err);
}
return attrbuf;
}
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
static void genl_family_rcv_msg_attrs_free(struct nlattr **attrbuf)
{
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
kfree(attrbuf);
}
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
struct genl_start_context {
const struct genl_family *family;
struct nlmsghdr *nlh;
struct netlink_ext_ack *extack;
const struct genl_split_ops *ops;
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
int hdrlen;
};
static int genl_start(struct netlink_callback *cb)
{
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
struct genl_start_context *ctx = cb->data;
const struct genl_split_ops *ops;
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
struct genl_dumpit_info *info;
struct nlattr **attrs = NULL;
int rc = 0;
ops = ctx->ops;
if (!(ops->validate & GENL_DONT_VALIDATE_DUMP) &&
ctx->nlh->nlmsg_len < nlmsg_msg_size(ctx->hdrlen))
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
return -EINVAL;
attrs = genl_family_rcv_msg_attrs_parse(ctx->family, ctx->nlh, ctx->extack,
ops, ctx->hdrlen,
GENL_DONT_VALIDATE_DUMP_STRICT);
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
if (IS_ERR(attrs))
return PTR_ERR(attrs);
info = genl_dumpit_info_alloc();
if (!info) {
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
genl_family_rcv_msg_attrs_free(attrs);
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
return -ENOMEM;
}
info->family = ctx->family;
info->op = *ops;
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
info->attrs = attrs;
cb->data = info;
if (ops->start) {
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
if (!ctx->family->parallel_ops)
genl_lock();
rc = ops->start(cb);
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
if (!ctx->family->parallel_ops)
genl_unlock();
}
if (rc) {
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
genl_family_rcv_msg_attrs_free(info->attrs);
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
genl_dumpit_info_free(info);
cb->data = NULL;
}
return rc;
}
static int genl_lock_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
{
const struct genl_split_ops *ops = &genl_dumpit_info(cb)->op;
int rc;
genl_lock();
rc = ops->dumpit(skb, cb);
genl_unlock();
return rc;
}
static int genl_lock_done(struct netlink_callback *cb)
{
const struct genl_dumpit_info *info = genl_dumpit_info(cb);
const struct genl_split_ops *ops = &info->op;
int rc = 0;
if (ops->done) {
genl_lock();
rc = ops->done(cb);
genl_unlock();
}
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
genl_family_rcv_msg_attrs_free(info->attrs);
genl_dumpit_info_free(info);
return rc;
}
static int genl_parallel_done(struct netlink_callback *cb)
{
const struct genl_dumpit_info *info = genl_dumpit_info(cb);
const struct genl_split_ops *ops = &info->op;
int rc = 0;
if (ops->done)
rc = ops->done(cb);
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
genl_family_rcv_msg_attrs_free(info->attrs);
genl_dumpit_info_free(info);
return rc;
}
static int genl_family_rcv_msg_dumpit(const struct genl_family *family,
struct sk_buff *skb,
struct nlmsghdr *nlh,
struct netlink_ext_ack *extack,
const struct genl_split_ops *ops,
int hdrlen, struct net *net)
{
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
struct genl_start_context ctx;
int err;
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
ctx.family = family;
ctx.nlh = nlh;
ctx.extack = extack;
ctx.ops = ops;
ctx.hdrlen = hdrlen;
if (!family->parallel_ops) {
struct netlink_dump_control c = {
.module = family->module,
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
.data = &ctx,
.start = genl_start,
.dump = genl_lock_dumpit,
.done = genl_lock_done,
};
genl_unlock();
err = __netlink_dump_start(net->genl_sock, skb, nlh, &c);
genl_lock();
} else {
struct netlink_dump_control c = {
.module = family->module,
genetlink: fix memory leaks in genl_family_rcv_msg_dumpit() There are two kinds of memory leaks in genl_family_rcv_msg_dumpit(): 1. Before we call ops->start(), whenever an error happens, we forget to free the memory allocated in genl_family_rcv_msg_dumpit(). 2. When ops->start() fails, the 'info' has been already installed on the per socket control block, so we should not free it here. More importantly, nlk->cb_running is still false at this point, so netlink_sock_destruct() cannot free it either. The first kind of memory leaks is easier to resolve, but the second one requires some deeper thoughts. After reviewing how netfilter handles this, the most elegant solution I find is just to use a similar way to allocate the memory, that is, moving memory allocations from caller into ops->start(). With this, we can solve both kinds of memory leaks: for 1), no memory allocation happens before ops->start(); for 2), ops->start() handles its own failures and 'info' is installed to the socket control block only when success. The only ugliness here is we have to pass all local variables on stack via a struct, but this is not hard to understand. Alternatively, we can introduce a ops->free() to solve this too, but it is overkill as only genetlink has this problem so far. Fixes: 1927f41a22a0 ("net: genetlink: introduce dump info struct to be available during dumpit op") Reported-by: syzbot+21f04f481f449c8db840@syzkaller.appspotmail.com Cc: "Jason A. Donenfeld" <Jason@zx2c4.com> Cc: Florian Westphal <fw@strlen.de> Cc: Pablo Neira Ayuso <pablo@netfilter.org> Cc: Jiri Pirko <jiri@mellanox.com> Cc: YueHaibing <yuehaibing@huawei.com> Cc: Shaochun Chen <cscnull@gmail.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-03 07:49:10 +03:00
.data = &ctx,
.start = genl_start,
.dump = ops->dumpit,
.done = genl_parallel_done,
};
err = __netlink_dump_start(net->genl_sock, skb, nlh, &c);
}
return err;
}
static int genl_family_rcv_msg_doit(const struct genl_family *family,
struct sk_buff *skb,
struct nlmsghdr *nlh,
struct netlink_ext_ack *extack,
const struct genl_split_ops *ops,
int hdrlen, struct net *net)
{
struct nlattr **attrbuf;
struct genl_info info;
int err;
attrbuf = genl_family_rcv_msg_attrs_parse(family, nlh, extack,
ops, hdrlen,
GENL_DONT_VALIDATE_STRICT);
if (IS_ERR(attrbuf))
return PTR_ERR(attrbuf);
info.snd_seq = nlh->nlmsg_seq;
info.snd_portid = NETLINK_CB(skb).portid;
info.nlhdr = nlh;
info.genlhdr = nlmsg_data(nlh);
info.userhdr = nlmsg_data(nlh) + GENL_HDRLEN;
info.attrs = attrbuf;
info.extack = extack;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
genl_info_net_set(&info, net);
memset(&info.user_ptr, 0, sizeof(info.user_ptr));
if (ops->pre_doit) {
err = ops->pre_doit(ops, skb, &info);
if (err)
goto out;
}
err = ops->doit(skb, &info);
if (ops->post_doit)
ops->post_doit(ops, skb, &info);
out:
genetlink: get rid of family->attrbuf genl_family_rcv_msg_attrs_parse() reuses the global family->attrbuf when family->parallel_ops is false. However, family->attrbuf is not protected by any lock on the genl_family_rcv_msg_doit() code path. This leads to several different consequences, one of them is UAF, like the following: genl_family_rcv_msg_doit(): genl_start(): genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); genl_family_rcv_msg_attrs_parse() attrbuf = family->attrbuf __nlmsg_parse(attrbuf); info->attrs = attrs; cb->data = info; netlink_unicast_kernel(): consume_skb() genl_lock_dumpit(): genl_dumpit_info(cb)->attrs Note family->attrbuf is an array of pointers to the skb data, once the skb is freed, any dereference of family->attrbuf will be a UAF. Maybe we could serialize the family->attrbuf with genl_mutex too, but that would make the locking more complicated. Instead, we can just get rid of family->attrbuf and always allocate attrbuf from heap like the family->parallel_ops==true code path. This may add some performance overhead but comparing with taking the global genl_mutex, it still looks better. Fixes: 75cdbdd08900 ("net: ieee802154: have genetlink code to parse the attrs during dumpit") Fixes: 057af7071344 ("net: tipc: have genetlink code to parse the attrs during dumpit") Reported-and-tested-by: syzbot+3039ddf6d7b13daf3787@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+80cad1e3cb4c41cde6ff@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+736bcbcb11b60d0c0792@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+520f8704db2b68091d44@syzkaller.appspotmail.com Reported-and-tested-by: syzbot+c96e4dfb32f8987fdeed@syzkaller.appspotmail.com Cc: Jiri Pirko <jiri@mellanox.com> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-06-27 10:12:24 +03:00
genl_family_rcv_msg_attrs_free(attrbuf);
return err;
}
static int genl_header_check(const struct genl_family *family,
struct nlmsghdr *nlh, struct genlmsghdr *hdr,
struct netlink_ext_ack *extack)
{
u16 flags;
/* Only for commands added after we started validating */
if (hdr->cmd < family->resv_start_op)
return 0;
if (hdr->reserved) {
NL_SET_ERR_MSG(extack, "genlmsghdr.reserved field is not 0");
return -EINVAL;
}
/* Old netlink flags have pretty loose semantics, allow only the flags
* consumed by the core where we can enforce the meaning.
*/
flags = nlh->nlmsg_flags;
if ((flags & NLM_F_DUMP) == NLM_F_DUMP) /* DUMP is 2 bits */
flags &= ~NLM_F_DUMP;
if (flags & ~(NLM_F_REQUEST | NLM_F_ACK | NLM_F_ECHO)) {
NL_SET_ERR_MSG(extack,
"ambiguous or reserved bits set in nlmsg_flags");
return -EINVAL;
}
return 0;
}
static int genl_family_rcv_msg(const struct genl_family *family,
struct sk_buff *skb,
struct nlmsghdr *nlh,
struct netlink_ext_ack *extack)
{
struct net *net = sock_net(skb->sk);
struct genlmsghdr *hdr = nlmsg_data(nlh);
struct genl_split_ops op;
int hdrlen;
u8 flags;
/* this family doesn't exist in this netns */
if (!family->netnsok && !net_eq(net, &init_net))
return -ENOENT;
hdrlen = GENL_HDRLEN + family->hdrsize;
if (nlh->nlmsg_len < nlmsg_msg_size(hdrlen))
return -EINVAL;
if (genl_header_check(family, nlh, hdr, extack))
return -EINVAL;
flags = (nlh->nlmsg_flags & NLM_F_DUMP) == NLM_F_DUMP ?
GENL_CMD_CAP_DUMP : GENL_CMD_CAP_DO;
if (genl_get_cmd(hdr->cmd, flags, family, &op))
return -EOPNOTSUPP;
if ((op.flags & GENL_ADMIN_PERM) &&
!netlink_capable(skb, CAP_NET_ADMIN))
return -EPERM;
if ((op.flags & GENL_UNS_ADMIN_PERM) &&
!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
return -EPERM;
if (flags & GENL_CMD_CAP_DUMP)
return genl_family_rcv_msg_dumpit(family, skb, nlh, extack,
&op, hdrlen, net);
else
return genl_family_rcv_msg_doit(family, skb, nlh, extack,
&op, hdrlen, net);
}
static int genl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
struct netlink_ext_ack *extack)
{
const struct genl_family *family;
int err;
family = genl_family_find_byid(nlh->nlmsg_type);
if (family == NULL)
return -ENOENT;
if (!family->parallel_ops)
genl_lock();
err = genl_family_rcv_msg(family, skb, nlh, extack);
if (!family->parallel_ops)
genl_unlock();
return err;
}
static void genl_rcv(struct sk_buff *skb)
{
down_read(&cb_lock);
netlink_rcv_skb(skb, &genl_rcv_msg);
up_read(&cb_lock);
}
/**************************************************************************
* Controller
**************************************************************************/
static struct genl_family genl_ctrl;
static int ctrl_fill_info(const struct genl_family *family, u32 portid, u32 seq,
u32 flags, struct sk_buff *skb, u8 cmd)
{
struct genl_op_iter i;
void *hdr;
hdr = genlmsg_put(skb, portid, seq, &genl_ctrl, flags, cmd);
if (hdr == NULL)
return -1;
if (nla_put_string(skb, CTRL_ATTR_FAMILY_NAME, family->name) ||
nla_put_u16(skb, CTRL_ATTR_FAMILY_ID, family->id) ||
nla_put_u32(skb, CTRL_ATTR_VERSION, family->version) ||
nla_put_u32(skb, CTRL_ATTR_HDRSIZE, family->hdrsize) ||
nla_put_u32(skb, CTRL_ATTR_MAXATTR, family->maxattr))
goto nla_put_failure;
if (genl_op_iter_init(family, &i)) {
struct nlattr *nla_ops;
nla_ops = nla_nest_start_noflag(skb, CTRL_ATTR_OPS);
if (nla_ops == NULL)
goto nla_put_failure;
while (genl_op_iter_next(&i)) {
struct nlattr *nest;
u32 op_flags;
op_flags = i.flags;
if (i.doit.policy || i.dumpit.policy)
op_flags |= GENL_CMD_CAP_HASPOL;
nest = nla_nest_start_noflag(skb, genl_op_iter_idx(&i));
if (nest == NULL)
goto nla_put_failure;
if (nla_put_u32(skb, CTRL_ATTR_OP_ID, i.cmd) ||
nla_put_u32(skb, CTRL_ATTR_OP_FLAGS, op_flags))
goto nla_put_failure;
nla_nest_end(skb, nest);
}
nla_nest_end(skb, nla_ops);
}
if (family->n_mcgrps) {
struct nlattr *nla_grps;
int i;
nla_grps = nla_nest_start_noflag(skb, CTRL_ATTR_MCAST_GROUPS);
if (nla_grps == NULL)
goto nla_put_failure;
for (i = 0; i < family->n_mcgrps; i++) {
struct nlattr *nest;
const struct genl_multicast_group *grp;
grp = &family->mcgrps[i];
nest = nla_nest_start_noflag(skb, i + 1);
if (nest == NULL)
goto nla_put_failure;
if (nla_put_u32(skb, CTRL_ATTR_MCAST_GRP_ID,
family->mcgrp_offset + i) ||
nla_put_string(skb, CTRL_ATTR_MCAST_GRP_NAME,
grp->name))
goto nla_put_failure;
nla_nest_end(skb, nest);
}
nla_nest_end(skb, nla_grps);
}
netlink: make nlmsg_end() and genlmsg_end() void Contrary to common expectations for an "int" return, these functions return only a positive value -- if used correctly they cannot even return 0 because the message header will necessarily be in the skb. This makes the very common pattern of if (genlmsg_end(...) < 0) { ... } be a whole bunch of dead code. Many places also simply do return nlmsg_end(...); and the caller is expected to deal with it. This also commonly (at least for me) causes errors, because it is very common to write if (my_function(...)) /* error condition */ and if my_function() does "return nlmsg_end()" this is of course wrong. Additionally, there's not a single place in the kernel that actually needs the message length returned, and if anyone needs it later then it'll be very easy to just use skb->len there. Remove this, and make the functions void. This removes a bunch of dead code as described above. The patch adds lines because I did - return nlmsg_end(...); + nlmsg_end(...); + return 0; I could have preserved all the function's return values by returning skb->len, but instead I've audited all the places calling the affected functions and found that none cared. A few places actually compared the return value with <= 0 in dump functionality, but that could just be changed to < 0 with no change in behaviour, so I opted for the more efficient version. One instance of the error I've made numerous times now is also present in net/phonet/pn_netlink.c in the route_dumpit() function - it didn't check for <0 or <=0 and thus broke out of the loop every single time. I've preserved this since it will (I think) have caused the messages to userspace to be formatted differently with just a single message for every SKB returned to userspace. It's possible that this isn't needed for the tools that actually use this, but I don't even know what they are so couldn't test that changing this behaviour would be acceptable. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-17 00:09:00 +03:00
genlmsg_end(skb, hdr);
return 0;
nla_put_failure:
genlmsg_cancel(skb, hdr);
return -EMSGSIZE;
}
static int ctrl_fill_mcgrp_info(const struct genl_family *family,
const struct genl_multicast_group *grp,
int grp_id, u32 portid, u32 seq, u32 flags,
struct sk_buff *skb, u8 cmd)
{
void *hdr;
struct nlattr *nla_grps;
struct nlattr *nest;
hdr = genlmsg_put(skb, portid, seq, &genl_ctrl, flags, cmd);
if (hdr == NULL)
return -1;
if (nla_put_string(skb, CTRL_ATTR_FAMILY_NAME, family->name) ||
nla_put_u16(skb, CTRL_ATTR_FAMILY_ID, family->id))
goto nla_put_failure;
nla_grps = nla_nest_start_noflag(skb, CTRL_ATTR_MCAST_GROUPS);
if (nla_grps == NULL)
goto nla_put_failure;
nest = nla_nest_start_noflag(skb, 1);
if (nest == NULL)
goto nla_put_failure;
if (nla_put_u32(skb, CTRL_ATTR_MCAST_GRP_ID, grp_id) ||
nla_put_string(skb, CTRL_ATTR_MCAST_GRP_NAME,
grp->name))
goto nla_put_failure;
nla_nest_end(skb, nest);
nla_nest_end(skb, nla_grps);
netlink: make nlmsg_end() and genlmsg_end() void Contrary to common expectations for an "int" return, these functions return only a positive value -- if used correctly they cannot even return 0 because the message header will necessarily be in the skb. This makes the very common pattern of if (genlmsg_end(...) < 0) { ... } be a whole bunch of dead code. Many places also simply do return nlmsg_end(...); and the caller is expected to deal with it. This also commonly (at least for me) causes errors, because it is very common to write if (my_function(...)) /* error condition */ and if my_function() does "return nlmsg_end()" this is of course wrong. Additionally, there's not a single place in the kernel that actually needs the message length returned, and if anyone needs it later then it'll be very easy to just use skb->len there. Remove this, and make the functions void. This removes a bunch of dead code as described above. The patch adds lines because I did - return nlmsg_end(...); + nlmsg_end(...); + return 0; I could have preserved all the function's return values by returning skb->len, but instead I've audited all the places calling the affected functions and found that none cared. A few places actually compared the return value with <= 0 in dump functionality, but that could just be changed to < 0 with no change in behaviour, so I opted for the more efficient version. One instance of the error I've made numerous times now is also present in net/phonet/pn_netlink.c in the route_dumpit() function - it didn't check for <0 or <=0 and thus broke out of the loop every single time. I've preserved this since it will (I think) have caused the messages to userspace to be formatted differently with just a single message for every SKB returned to userspace. It's possible that this isn't needed for the tools that actually use this, but I don't even know what they are so couldn't test that changing this behaviour would be acceptable. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-01-17 00:09:00 +03:00
genlmsg_end(skb, hdr);
return 0;
nla_put_failure:
genlmsg_cancel(skb, hdr);
return -EMSGSIZE;
}
static int ctrl_dumpfamily(struct sk_buff *skb, struct netlink_callback *cb)
{
int n = 0;
struct genl_family *rt;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
struct net *net = sock_net(skb->sk);
int fams_to_skip = cb->args[0];
unsigned int id;
idr_for_each_entry(&genl_fam_idr, rt, id) {
if (!rt->netnsok && !net_eq(net, &init_net))
continue;
if (n++ < fams_to_skip)
continue;
if (ctrl_fill_info(rt, NETLINK_CB(cb->skb).portid,
cb->nlh->nlmsg_seq, NLM_F_MULTI,
skb, CTRL_CMD_NEWFAMILY) < 0) {
n--;
break;
}
}
cb->args[0] = n;
return skb->len;
}
static struct sk_buff *ctrl_build_family_msg(const struct genl_family *family,
u32 portid, int seq, u8 cmd)
{
struct sk_buff *skb;
int err;
skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
if (skb == NULL)
return ERR_PTR(-ENOBUFS);
err = ctrl_fill_info(family, portid, seq, 0, skb, cmd);
if (err < 0) {
nlmsg_free(skb);
return ERR_PTR(err);
}
return skb;
}
static struct sk_buff *
ctrl_build_mcgrp_msg(const struct genl_family *family,
const struct genl_multicast_group *grp,
int grp_id, u32 portid, int seq, u8 cmd)
{
struct sk_buff *skb;
int err;
skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
if (skb == NULL)
return ERR_PTR(-ENOBUFS);
err = ctrl_fill_mcgrp_info(family, grp, grp_id, portid,
seq, 0, skb, cmd);
if (err < 0) {
nlmsg_free(skb);
return ERR_PTR(err);
}
return skb;
}
static const struct nla_policy ctrl_policy_family[] = {
[CTRL_ATTR_FAMILY_ID] = { .type = NLA_U16 },
[CTRL_ATTR_FAMILY_NAME] = { .type = NLA_NUL_STRING,
.len = GENL_NAMSIZ - 1 },
};
static int ctrl_getfamily(struct sk_buff *skb, struct genl_info *info)
{
struct sk_buff *msg;
const struct genl_family *res = NULL;
int err = -EINVAL;
if (info->attrs[CTRL_ATTR_FAMILY_ID]) {
u16 id = nla_get_u16(info->attrs[CTRL_ATTR_FAMILY_ID]);
res = genl_family_find_byid(id);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
err = -ENOENT;
}
if (info->attrs[CTRL_ATTR_FAMILY_NAME]) {
char *name;
name = nla_data(info->attrs[CTRL_ATTR_FAMILY_NAME]);
res = genl_family_find_byname(name);
#ifdef CONFIG_MODULES
if (res == NULL) {
genl_unlock();
genetlink: release cb_lock before requesting additional module Requesting external module with cb_lock taken can result in the deadlock like showed below: [ 2458.111347] Showing all locks held in the system: [ 2458.111347] 1 lock held by NetworkManager/582: [ 2458.111347] #0: (cb_lock){++++++}, at: [<ffffffff8162bc79>] genl_rcv+0x19/0x40 [ 2458.111347] 1 lock held by modprobe/603: [ 2458.111347] #0: (cb_lock){++++++}, at: [<ffffffff8162baa5>] genl_lock_all+0x15/0x30 [ 2461.579457] SysRq : Show Blocked State [ 2461.580103] task PC stack pid father [ 2461.580103] NetworkManager D ffff880034b84500 4040 582 1 0x00000080 [ 2461.580103] ffff8800197ff720 0000000000000046 00000000001d5340 ffff8800197fffd8 [ 2461.580103] ffff8800197fffd8 00000000001d5340 ffff880019631700 7fffffffffffffff [ 2461.580103] ffff8800197ff880 ffff8800197ff878 ffff880019631700 ffff880019631700 [ 2461.580103] Call Trace: [ 2461.580103] [<ffffffff817355f9>] schedule+0x29/0x70 [ 2461.580103] [<ffffffff81731ad1>] schedule_timeout+0x1c1/0x360 [ 2461.580103] [<ffffffff810e69eb>] ? mark_held_locks+0xbb/0x140 [ 2461.580103] [<ffffffff817377ac>] ? _raw_spin_unlock_irq+0x2c/0x50 [ 2461.580103] [<ffffffff810e6b6d>] ? trace_hardirqs_on_caller+0xfd/0x1c0 [ 2461.580103] [<ffffffff81736398>] wait_for_completion_killable+0xe8/0x170 [ 2461.580103] [<ffffffff810b7fa0>] ? wake_up_state+0x20/0x20 [ 2461.580103] [<ffffffff81095825>] call_usermodehelper_exec+0x1a5/0x210 [ 2461.580103] [<ffffffff817362ed>] ? wait_for_completion_killable+0x3d/0x170 [ 2461.580103] [<ffffffff81095cc3>] __request_module+0x1b3/0x370 [ 2461.580103] [<ffffffff810e6b6d>] ? trace_hardirqs_on_caller+0xfd/0x1c0 [ 2461.580103] [<ffffffff8162c5c9>] ctrl_getfamily+0x159/0x190 [ 2461.580103] [<ffffffff8162d8a4>] genl_family_rcv_msg+0x1f4/0x2e0 [ 2461.580103] [<ffffffff8162d990>] ? genl_family_rcv_msg+0x2e0/0x2e0 [ 2461.580103] [<ffffffff8162da1e>] genl_rcv_msg+0x8e/0xd0 [ 2461.580103] [<ffffffff8162b729>] netlink_rcv_skb+0xa9/0xc0 [ 2461.580103] [<ffffffff8162bc88>] genl_rcv+0x28/0x40 [ 2461.580103] [<ffffffff8162ad6d>] netlink_unicast+0xdd/0x190 [ 2461.580103] [<ffffffff8162b149>] netlink_sendmsg+0x329/0x750 [ 2461.580103] [<ffffffff815db849>] sock_sendmsg+0x99/0xd0 [ 2461.580103] [<ffffffff810bb58f>] ? local_clock+0x5f/0x70 [ 2461.580103] [<ffffffff810e96e8>] ? lock_release_non_nested+0x308/0x350 [ 2461.580103] [<ffffffff815dbc6e>] ___sys_sendmsg+0x39e/0x3b0 [ 2461.580103] [<ffffffff810565af>] ? kvm_clock_read+0x2f/0x50 [ 2461.580103] [<ffffffff810218b9>] ? sched_clock+0x9/0x10 [ 2461.580103] [<ffffffff810bb2bd>] ? sched_clock_local+0x1d/0x80 [ 2461.580103] [<ffffffff810bb448>] ? sched_clock_cpu+0xa8/0x100 [ 2461.580103] [<ffffffff810e33ad>] ? trace_hardirqs_off+0xd/0x10 [ 2461.580103] [<ffffffff810bb58f>] ? local_clock+0x5f/0x70 [ 2461.580103] [<ffffffff810e3f7f>] ? lock_release_holdtime.part.28+0xf/0x1a0 [ 2461.580103] [<ffffffff8120fec9>] ? fget_light+0xf9/0x510 [ 2461.580103] [<ffffffff8120fe0c>] ? fget_light+0x3c/0x510 [ 2461.580103] [<ffffffff815dd1d2>] __sys_sendmsg+0x42/0x80 [ 2461.580103] [<ffffffff815dd222>] SyS_sendmsg+0x12/0x20 [ 2461.580103] [<ffffffff81741ad9>] system_call_fastpath+0x16/0x1b [ 2461.580103] modprobe D ffff88000f2c8000 4632 603 602 0x00000080 [ 2461.580103] ffff88000f04fba8 0000000000000046 00000000001d5340 ffff88000f04ffd8 [ 2461.580103] ffff88000f04ffd8 00000000001d5340 ffff8800377d4500 ffff8800377d4500 [ 2461.580103] ffffffff81d0b260 ffffffff81d0b268 ffffffff00000000 ffffffff81d0b2b0 [ 2461.580103] Call Trace: [ 2461.580103] [<ffffffff817355f9>] schedule+0x29/0x70 [ 2461.580103] [<ffffffff81736d4d>] rwsem_down_write_failed+0xed/0x1a0 [ 2461.580103] [<ffffffff810bb200>] ? update_cpu_load_active+0x10/0xb0 [ 2461.580103] [<ffffffff8137b473>] call_rwsem_down_write_failed+0x13/0x20 [ 2461.580103] [<ffffffff8173492d>] ? down_write+0x9d/0xb2 [ 2461.580103] [<ffffffff8162baa5>] ? genl_lock_all+0x15/0x30 [ 2461.580103] [<ffffffff8162baa5>] genl_lock_all+0x15/0x30 [ 2461.580103] [<ffffffff8162cbb3>] genl_register_family+0x53/0x1f0 [ 2461.580103] [<ffffffffa01dc000>] ? 0xffffffffa01dbfff [ 2461.580103] [<ffffffff8162d650>] genl_register_family_with_ops+0x20/0x80 [ 2461.580103] [<ffffffffa01dc000>] ? 0xffffffffa01dbfff [ 2461.580103] [<ffffffffa017fe84>] nl80211_init+0x24/0xf0 [cfg80211] [ 2461.580103] [<ffffffffa01dc000>] ? 0xffffffffa01dbfff [ 2461.580103] [<ffffffffa01dc043>] cfg80211_init+0x43/0xdb [cfg80211] [ 2461.580103] [<ffffffff810020fa>] do_one_initcall+0xfa/0x1b0 [ 2461.580103] [<ffffffff8105cb93>] ? set_memory_nx+0x43/0x50 [ 2461.580103] [<ffffffff810f75af>] load_module+0x1c6f/0x27f0 [ 2461.580103] [<ffffffff810f2c90>] ? store_uevent+0x40/0x40 [ 2461.580103] [<ffffffff810f82c6>] SyS_finit_module+0x86/0xb0 [ 2461.580103] [<ffffffff81741ad9>] system_call_fastpath+0x16/0x1b [ 2461.580103] Sched Debug Version: v0.10, 3.11.0-0.rc1.git4.1.fc20.x86_64 #1 Problem start to happen after adding net-pf-16-proto-16-family-nl80211 alias name to cfg80211 module by below commit (though that commit itself is perfectly fine): commit fb4e156886ce6e8309e912d8b370d192330d19d3 Author: Marcel Holtmann <marcel@holtmann.org> Date: Sun Apr 28 16:22:06 2013 -0700 nl80211: Add generic netlink module alias for cfg80211/nl80211 Reported-and-tested-by: Jeff Layton <jlayton@redhat.com> Reported-by: Richard W.M. Jones <rjones@redhat.com> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com> Reviewed-by: Pravin B Shelar <pshelar@nicira.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-26 13:00:10 +04:00
up_read(&cb_lock);
request_module("net-pf-%d-proto-%d-family-%s",
PF_NETLINK, NETLINK_GENERIC, name);
genetlink: release cb_lock before requesting additional module Requesting external module with cb_lock taken can result in the deadlock like showed below: [ 2458.111347] Showing all locks held in the system: [ 2458.111347] 1 lock held by NetworkManager/582: [ 2458.111347] #0: (cb_lock){++++++}, at: [<ffffffff8162bc79>] genl_rcv+0x19/0x40 [ 2458.111347] 1 lock held by modprobe/603: [ 2458.111347] #0: (cb_lock){++++++}, at: [<ffffffff8162baa5>] genl_lock_all+0x15/0x30 [ 2461.579457] SysRq : Show Blocked State [ 2461.580103] task PC stack pid father [ 2461.580103] NetworkManager D ffff880034b84500 4040 582 1 0x00000080 [ 2461.580103] ffff8800197ff720 0000000000000046 00000000001d5340 ffff8800197fffd8 [ 2461.580103] ffff8800197fffd8 00000000001d5340 ffff880019631700 7fffffffffffffff [ 2461.580103] ffff8800197ff880 ffff8800197ff878 ffff880019631700 ffff880019631700 [ 2461.580103] Call Trace: [ 2461.580103] [<ffffffff817355f9>] schedule+0x29/0x70 [ 2461.580103] [<ffffffff81731ad1>] schedule_timeout+0x1c1/0x360 [ 2461.580103] [<ffffffff810e69eb>] ? mark_held_locks+0xbb/0x140 [ 2461.580103] [<ffffffff817377ac>] ? _raw_spin_unlock_irq+0x2c/0x50 [ 2461.580103] [<ffffffff810e6b6d>] ? trace_hardirqs_on_caller+0xfd/0x1c0 [ 2461.580103] [<ffffffff81736398>] wait_for_completion_killable+0xe8/0x170 [ 2461.580103] [<ffffffff810b7fa0>] ? wake_up_state+0x20/0x20 [ 2461.580103] [<ffffffff81095825>] call_usermodehelper_exec+0x1a5/0x210 [ 2461.580103] [<ffffffff817362ed>] ? wait_for_completion_killable+0x3d/0x170 [ 2461.580103] [<ffffffff81095cc3>] __request_module+0x1b3/0x370 [ 2461.580103] [<ffffffff810e6b6d>] ? trace_hardirqs_on_caller+0xfd/0x1c0 [ 2461.580103] [<ffffffff8162c5c9>] ctrl_getfamily+0x159/0x190 [ 2461.580103] [<ffffffff8162d8a4>] genl_family_rcv_msg+0x1f4/0x2e0 [ 2461.580103] [<ffffffff8162d990>] ? genl_family_rcv_msg+0x2e0/0x2e0 [ 2461.580103] [<ffffffff8162da1e>] genl_rcv_msg+0x8e/0xd0 [ 2461.580103] [<ffffffff8162b729>] netlink_rcv_skb+0xa9/0xc0 [ 2461.580103] [<ffffffff8162bc88>] genl_rcv+0x28/0x40 [ 2461.580103] [<ffffffff8162ad6d>] netlink_unicast+0xdd/0x190 [ 2461.580103] [<ffffffff8162b149>] netlink_sendmsg+0x329/0x750 [ 2461.580103] [<ffffffff815db849>] sock_sendmsg+0x99/0xd0 [ 2461.580103] [<ffffffff810bb58f>] ? local_clock+0x5f/0x70 [ 2461.580103] [<ffffffff810e96e8>] ? lock_release_non_nested+0x308/0x350 [ 2461.580103] [<ffffffff815dbc6e>] ___sys_sendmsg+0x39e/0x3b0 [ 2461.580103] [<ffffffff810565af>] ? kvm_clock_read+0x2f/0x50 [ 2461.580103] [<ffffffff810218b9>] ? sched_clock+0x9/0x10 [ 2461.580103] [<ffffffff810bb2bd>] ? sched_clock_local+0x1d/0x80 [ 2461.580103] [<ffffffff810bb448>] ? sched_clock_cpu+0xa8/0x100 [ 2461.580103] [<ffffffff810e33ad>] ? trace_hardirqs_off+0xd/0x10 [ 2461.580103] [<ffffffff810bb58f>] ? local_clock+0x5f/0x70 [ 2461.580103] [<ffffffff810e3f7f>] ? lock_release_holdtime.part.28+0xf/0x1a0 [ 2461.580103] [<ffffffff8120fec9>] ? fget_light+0xf9/0x510 [ 2461.580103] [<ffffffff8120fe0c>] ? fget_light+0x3c/0x510 [ 2461.580103] [<ffffffff815dd1d2>] __sys_sendmsg+0x42/0x80 [ 2461.580103] [<ffffffff815dd222>] SyS_sendmsg+0x12/0x20 [ 2461.580103] [<ffffffff81741ad9>] system_call_fastpath+0x16/0x1b [ 2461.580103] modprobe D ffff88000f2c8000 4632 603 602 0x00000080 [ 2461.580103] ffff88000f04fba8 0000000000000046 00000000001d5340 ffff88000f04ffd8 [ 2461.580103] ffff88000f04ffd8 00000000001d5340 ffff8800377d4500 ffff8800377d4500 [ 2461.580103] ffffffff81d0b260 ffffffff81d0b268 ffffffff00000000 ffffffff81d0b2b0 [ 2461.580103] Call Trace: [ 2461.580103] [<ffffffff817355f9>] schedule+0x29/0x70 [ 2461.580103] [<ffffffff81736d4d>] rwsem_down_write_failed+0xed/0x1a0 [ 2461.580103] [<ffffffff810bb200>] ? update_cpu_load_active+0x10/0xb0 [ 2461.580103] [<ffffffff8137b473>] call_rwsem_down_write_failed+0x13/0x20 [ 2461.580103] [<ffffffff8173492d>] ? down_write+0x9d/0xb2 [ 2461.580103] [<ffffffff8162baa5>] ? genl_lock_all+0x15/0x30 [ 2461.580103] [<ffffffff8162baa5>] genl_lock_all+0x15/0x30 [ 2461.580103] [<ffffffff8162cbb3>] genl_register_family+0x53/0x1f0 [ 2461.580103] [<ffffffffa01dc000>] ? 0xffffffffa01dbfff [ 2461.580103] [<ffffffff8162d650>] genl_register_family_with_ops+0x20/0x80 [ 2461.580103] [<ffffffffa01dc000>] ? 0xffffffffa01dbfff [ 2461.580103] [<ffffffffa017fe84>] nl80211_init+0x24/0xf0 [cfg80211] [ 2461.580103] [<ffffffffa01dc000>] ? 0xffffffffa01dbfff [ 2461.580103] [<ffffffffa01dc043>] cfg80211_init+0x43/0xdb [cfg80211] [ 2461.580103] [<ffffffff810020fa>] do_one_initcall+0xfa/0x1b0 [ 2461.580103] [<ffffffff8105cb93>] ? set_memory_nx+0x43/0x50 [ 2461.580103] [<ffffffff810f75af>] load_module+0x1c6f/0x27f0 [ 2461.580103] [<ffffffff810f2c90>] ? store_uevent+0x40/0x40 [ 2461.580103] [<ffffffff810f82c6>] SyS_finit_module+0x86/0xb0 [ 2461.580103] [<ffffffff81741ad9>] system_call_fastpath+0x16/0x1b [ 2461.580103] Sched Debug Version: v0.10, 3.11.0-0.rc1.git4.1.fc20.x86_64 #1 Problem start to happen after adding net-pf-16-proto-16-family-nl80211 alias name to cfg80211 module by below commit (though that commit itself is perfectly fine): commit fb4e156886ce6e8309e912d8b370d192330d19d3 Author: Marcel Holtmann <marcel@holtmann.org> Date: Sun Apr 28 16:22:06 2013 -0700 nl80211: Add generic netlink module alias for cfg80211/nl80211 Reported-and-tested-by: Jeff Layton <jlayton@redhat.com> Reported-by: Richard W.M. Jones <rjones@redhat.com> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com> Reviewed-by: Pravin B Shelar <pshelar@nicira.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-07-26 13:00:10 +04:00
down_read(&cb_lock);
genl_lock();
res = genl_family_find_byname(name);
}
#endif
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
err = -ENOENT;
}
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
if (res == NULL)
return err;
if (!res->netnsok && !net_eq(genl_info_net(info), &init_net)) {
/* family doesn't exist here */
return -ENOENT;
}
msg = ctrl_build_family_msg(res, info->snd_portid, info->snd_seq,
CTRL_CMD_NEWFAMILY);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
if (IS_ERR(msg))
return PTR_ERR(msg);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
return genlmsg_reply(msg, info);
}
static int genl_ctrl_event(int event, const struct genl_family *family,
const struct genl_multicast_group *grp,
int grp_id)
{
struct sk_buff *msg;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
/* genl is still initialising */
if (!init_net.genl_sock)
return 0;
switch (event) {
case CTRL_CMD_NEWFAMILY:
case CTRL_CMD_DELFAMILY:
WARN_ON(grp);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
msg = ctrl_build_family_msg(family, 0, 0, event);
break;
case CTRL_CMD_NEWMCAST_GRP:
case CTRL_CMD_DELMCAST_GRP:
BUG_ON(!grp);
msg = ctrl_build_mcgrp_msg(family, grp, grp_id, 0, 0, event);
break;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
default:
return -EINVAL;
}
if (IS_ERR(msg))
return PTR_ERR(msg);
if (!family->netnsok) {
genlmsg_multicast_netns(&genl_ctrl, &init_net, msg, 0,
0, GFP_KERNEL);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
} else {
rcu_read_lock();
genlmsg_multicast_allns(&genl_ctrl, msg, 0,
0, GFP_ATOMIC);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
rcu_read_unlock();
}
return 0;
}
struct ctrl_dump_policy_ctx {
struct netlink_policy_dump_state *state;
const struct genl_family *rt;
struct genl_op_iter *op_iter;
u32 op;
u16 fam_id;
u8 dump_map:1,
single_op:1;
};
static const struct nla_policy ctrl_policy_policy[] = {
[CTRL_ATTR_FAMILY_ID] = { .type = NLA_U16 },
[CTRL_ATTR_FAMILY_NAME] = { .type = NLA_NUL_STRING,
.len = GENL_NAMSIZ - 1 },
[CTRL_ATTR_OP] = { .type = NLA_U32 },
};
static int ctrl_dumppolicy_start(struct netlink_callback *cb)
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
{
const struct genl_dumpit_info *info = genl_dumpit_info(cb);
struct ctrl_dump_policy_ctx *ctx = (void *)cb->ctx;
struct nlattr **tb = info->attrs;
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
const struct genl_family *rt;
struct genl_op_iter i;
int err;
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
BUILD_BUG_ON(sizeof(*ctx) > sizeof(cb->ctx));
if (!tb[CTRL_ATTR_FAMILY_ID] && !tb[CTRL_ATTR_FAMILY_NAME])
return -EINVAL;
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
if (tb[CTRL_ATTR_FAMILY_ID]) {
ctx->fam_id = nla_get_u16(tb[CTRL_ATTR_FAMILY_ID]);
} else {
rt = genl_family_find_byname(
nla_data(tb[CTRL_ATTR_FAMILY_NAME]));
if (!rt)
return -ENOENT;
ctx->fam_id = rt->id;
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
}
rt = genl_family_find_byid(ctx->fam_id);
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
if (!rt)
return -ENOENT;
ctx->rt = rt;
if (tb[CTRL_ATTR_OP]) {
struct genl_split_ops doit, dump;
ctx->single_op = true;
ctx->op = nla_get_u32(tb[CTRL_ATTR_OP]);
err = genl_get_cmd_both(ctx->op, rt, &doit, &dump);
if (err) {
NL_SET_BAD_ATTR(cb->extack, tb[CTRL_ATTR_OP]);
return err;
}
if (doit.policy) {
err = netlink_policy_dump_add_policy(&ctx->state,
doit.policy,
doit.maxattr);
if (err)
goto err_free_state;
}
if (dump.policy) {
err = netlink_policy_dump_add_policy(&ctx->state,
dump.policy,
dump.maxattr);
if (err)
goto err_free_state;
}
if (!ctx->state)
return -ENODATA;
ctx->dump_map = 1;
return 0;
}
ctx->op_iter = kmalloc(sizeof(*ctx->op_iter), GFP_KERNEL);
if (!ctx->op_iter)
return -ENOMEM;
genl_op_iter_init(rt, ctx->op_iter);
ctx->dump_map = genl_op_iter_next(ctx->op_iter);
for (genl_op_iter_init(rt, &i); genl_op_iter_next(&i); ) {
if (i.doit.policy) {
err = netlink_policy_dump_add_policy(&ctx->state,
i.doit.policy,
i.doit.maxattr);
if (err)
goto err_free_state;
}
if (i.dumpit.policy) {
err = netlink_policy_dump_add_policy(&ctx->state,
i.dumpit.policy,
i.dumpit.maxattr);
if (err)
goto err_free_state;
}
}
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
if (!ctx->state) {
err = -ENODATA;
goto err_free_op_iter;
}
return 0;
err_free_state:
netlink_policy_dump_free(ctx->state);
err_free_op_iter:
kfree(ctx->op_iter);
return err;
}
static void *ctrl_dumppolicy_prep(struct sk_buff *skb,
struct netlink_callback *cb)
{
struct ctrl_dump_policy_ctx *ctx = (void *)cb->ctx;
void *hdr;
hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid,
cb->nlh->nlmsg_seq, &genl_ctrl,
NLM_F_MULTI, CTRL_CMD_GETPOLICY);
if (!hdr)
return NULL;
if (nla_put_u16(skb, CTRL_ATTR_FAMILY_ID, ctx->fam_id))
return NULL;
return hdr;
}
static int ctrl_dumppolicy_put_op(struct sk_buff *skb,
struct netlink_callback *cb,
struct genl_split_ops *doit,
struct genl_split_ops *dumpit)
{
struct ctrl_dump_policy_ctx *ctx = (void *)cb->ctx;
struct nlattr *nest_pol, *nest_op;
void *hdr;
int idx;
/* skip if we have nothing to show */
if (!doit->policy && !dumpit->policy)
return 0;
hdr = ctrl_dumppolicy_prep(skb, cb);
if (!hdr)
return -ENOBUFS;
nest_pol = nla_nest_start(skb, CTRL_ATTR_OP_POLICY);
if (!nest_pol)
goto err;
nest_op = nla_nest_start(skb, doit->cmd);
if (!nest_op)
goto err;
if (doit->policy) {
idx = netlink_policy_dump_get_policy_idx(ctx->state,
doit->policy,
doit->maxattr);
if (nla_put_u32(skb, CTRL_ATTR_POLICY_DO, idx))
goto err;
}
if (dumpit->policy) {
idx = netlink_policy_dump_get_policy_idx(ctx->state,
dumpit->policy,
dumpit->maxattr);
if (nla_put_u32(skb, CTRL_ATTR_POLICY_DUMP, idx))
goto err;
}
nla_nest_end(skb, nest_op);
nla_nest_end(skb, nest_pol);
genlmsg_end(skb, hdr);
return 0;
err:
genlmsg_cancel(skb, hdr);
return -ENOBUFS;
}
static int ctrl_dumppolicy(struct sk_buff *skb, struct netlink_callback *cb)
{
struct ctrl_dump_policy_ctx *ctx = (void *)cb->ctx;
void *hdr;
if (ctx->dump_map) {
if (ctx->single_op) {
struct genl_split_ops doit, dumpit;
if (WARN_ON(genl_get_cmd_both(ctx->op, ctx->rt,
&doit, &dumpit)))
return -ENOENT;
if (ctrl_dumppolicy_put_op(skb, cb, &doit, &dumpit))
return skb->len;
/* done with the per-op policy index list */
ctx->dump_map = 0;
}
while (ctx->dump_map) {
if (ctrl_dumppolicy_put_op(skb, cb,
&ctx->op_iter->doit,
&ctx->op_iter->dumpit))
return skb->len;
ctx->dump_map = genl_op_iter_next(ctx->op_iter);
}
}
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
while (netlink_policy_dump_loop(ctx->state)) {
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
struct nlattr *nest;
hdr = ctrl_dumppolicy_prep(skb, cb);
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
if (!hdr)
goto nla_put_failure;
nest = nla_nest_start(skb, CTRL_ATTR_POLICY);
if (!nest)
goto nla_put_failure;
if (netlink_policy_dump_write(skb, ctx->state))
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
goto nla_put_failure;
nla_nest_end(skb, nest);
genlmsg_end(skb, hdr);
}
return skb->len;
nla_put_failure:
genlmsg_cancel(skb, hdr);
return skb->len;
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
}
static int ctrl_dumppolicy_done(struct netlink_callback *cb)
{
struct ctrl_dump_policy_ctx *ctx = (void *)cb->ctx;
kfree(ctx->op_iter);
netlink_policy_dump_free(ctx->state);
return 0;
}
static const struct genl_split_ops genl_ctrl_ops[] = {
{
.cmd = CTRL_CMD_GETFAMILY,
.validate = GENL_DONT_VALIDATE_STRICT,
.policy = ctrl_policy_family,
.maxattr = ARRAY_SIZE(ctrl_policy_family) - 1,
.doit = ctrl_getfamily,
.flags = GENL_CMD_CAP_DO,
},
{
.cmd = CTRL_CMD_GETFAMILY,
.validate = GENL_DONT_VALIDATE_DUMP,
.policy = ctrl_policy_family,
.maxattr = ARRAY_SIZE(ctrl_policy_family) - 1,
.dumpit = ctrl_dumpfamily,
.flags = GENL_CMD_CAP_DUMP,
},
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
{
.cmd = CTRL_CMD_GETPOLICY,
.policy = ctrl_policy_policy,
.maxattr = ARRAY_SIZE(ctrl_policy_policy) - 1,
.start = ctrl_dumppolicy_start,
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
.dumpit = ctrl_dumppolicy,
.done = ctrl_dumppolicy_done,
.flags = GENL_CMD_CAP_DUMP,
netlink: add infrastructure to expose policies to userspace Add, and use in generic netlink, helpers to dump out a netlink policy to userspace, including all the range validation data, nested policies etc. This lets userspace discover what the kernel understands. For families/commands other than generic netlink, the helpers need to be used directly in an appropriate command, or we can add some infrastructure (a new netlink family) that those can register their policies with for introspection. I'm not that familiar with non-generic netlink, so that's left out for now. The data exposed to userspace also includes min and max length for binary/string data, I've done that instead of letting the userspace tools figure out whether min/max is intended based on the type so that we can extend this later in the kernel, we might want to just use the range data for example. Because of this, I opted to not directly expose the NLA_* values, even if some of them are already exposed via BPF, as with min/max length we don't need to have different types here for NLA_BINARY/NLA_MIN_LEN/NLA_EXACT_LEN, we just make them all NL_ATTR_TYPE_BINARY with min/max length optionally set. Similarly, we don't really need NLA_MSECS, and perhaps can remove it in the future - but not if we encode it into the userspace API now. It gets mapped to NL_ATTR_TYPE_U64 here. Note that the exposing here corresponds to the strict policy interpretation, and NLA_UNSPEC items are omitted entirely. To get those, change them to NLA_MIN_LEN which behaves in exactly the same way, but is exposed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-04-30 23:13:12 +03:00
},
};
static const struct genl_multicast_group genl_ctrl_groups[] = {
{ .name = "notify", },
};
static struct genl_family genl_ctrl __ro_after_init = {
.module = THIS_MODULE,
.split_ops = genl_ctrl_ops,
.n_split_ops = ARRAY_SIZE(genl_ctrl_ops),
.resv_start_op = CTRL_CMD_GETPOLICY + 1,
.mcgrps = genl_ctrl_groups,
.n_mcgrps = ARRAY_SIZE(genl_ctrl_groups),
.id = GENL_ID_CTRL,
.name = "nlctrl",
.version = 0x2,
.netnsok = true,
};
static int genl_bind(struct net *net, int group)
{
const struct genl_family *family;
unsigned int id;
int ret = 0;
down_read(&cb_lock);
idr_for_each_entry(&genl_fam_idr, family, id) {
const struct genl_multicast_group *grp;
int i;
if (family->n_mcgrps == 0)
continue;
i = group - family->mcgrp_offset;
if (i < 0 || i >= family->n_mcgrps)
continue;
grp = &family->mcgrps[i];
if ((grp->flags & GENL_UNS_ADMIN_PERM) &&
!ns_capable(net->user_ns, CAP_NET_ADMIN))
ret = -EPERM;
break;
}
up_read(&cb_lock);
return ret;
}
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
static int __net_init genl_pernet_init(struct net *net)
{
struct netlink_kernel_cfg cfg = {
.input = genl_rcv,
.flags = NL_CFG_F_NONROOT_RECV,
.bind = genl_bind,
};
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
/* we'll bump the group number right afterwards */
net->genl_sock = netlink_kernel_create(net, NETLINK_GENERIC, &cfg);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
if (!net->genl_sock && net_eq(net, &init_net))
panic("GENL: Cannot initialize generic netlink\n");
if (!net->genl_sock)
return -ENOMEM;
return 0;
}
static void __net_exit genl_pernet_exit(struct net *net)
{
netlink_kernel_release(net->genl_sock);
net->genl_sock = NULL;
}
static struct pernet_operations genl_pernet_ops = {
.init = genl_pernet_init,
.exit = genl_pernet_exit,
};
static int __init genl_init(void)
{
int err;
err = genl_register_family(&genl_ctrl);
if (err < 0)
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
goto problem;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
err = register_pernet_subsys(&genl_pernet_ops);
if (err)
goto problem;
return 0;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
problem:
panic("GENL: Cannot register controller: %d\n", err);
}
core_initcall(genl_init);
static int genlmsg_mcast(struct sk_buff *skb, u32 portid, unsigned long group,
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
gfp_t flags)
{
struct sk_buff *tmp;
struct net *net, *prev = NULL;
bool delivered = false;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
int err;
for_each_net_rcu(net) {
if (prev) {
tmp = skb_clone(skb, flags);
if (!tmp) {
err = -ENOMEM;
goto error;
}
err = nlmsg_multicast(prev->genl_sock, tmp,
portid, group, flags);
if (!err)
delivered = true;
else if (err != -ESRCH)
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
goto error;
}
prev = net;
}
err = nlmsg_multicast(prev->genl_sock, skb, portid, group, flags);
if (!err)
delivered = true;
else if (err != -ESRCH)
return err;
return delivered ? 0 : -ESRCH;
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
error:
kfree_skb(skb);
return err;
}
int genlmsg_multicast_allns(const struct genl_family *family,
struct sk_buff *skb, u32 portid,
unsigned int group, gfp_t flags)
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
{
if (WARN_ON_ONCE(group >= family->n_mcgrps))
return -EINVAL;
group = family->mcgrp_offset + group;
return genlmsg_mcast(skb, portid, group, flags);
genetlink: make netns aware This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true. A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns). The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way. After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis. Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns. Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2009-07-10 13:51:34 +04:00
}
EXPORT_SYMBOL(genlmsg_multicast_allns);
void genl_notify(const struct genl_family *family, struct sk_buff *skb,
struct genl_info *info, u32 group, gfp_t flags)
{
struct net *net = genl_info_net(info);
struct sock *sk = net->genl_sock;
if (WARN_ON_ONCE(group >= family->n_mcgrps))
return;
group = family->mcgrp_offset + group;
nlmsg_notify(sk, skb, info->snd_portid, group,
nlmsg_report(info->nlhdr), flags);
}
EXPORT_SYMBOL(genl_notify);