iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
b805aa0e79
linux/include/asm-parisc/assembly.h

512 lines
12 KiB
C
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
/*
* Copyright (C) 1999 Hewlett-Packard (Frank Rowand)
* Copyright (C) 1999 Philipp Rumpf <prumpf@tux.org>
* Copyright (C) 1999 SuSE GmbH
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifndef _PARISC_ASSEMBLY_H
#define _PARISC_ASSEMBLY_H
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
#define CALLEE_FLOAT_FRAME_SIZE 80
[PARISC] Specify level to fix binutils level promotion bug fixup.S needs to specify .level and use correct LDREG macro. New binutils has a bug where it doesn't "promote" from PA1.0 to PA1.1 correctly when using ",s" completer. remove use of __LP64__ in assembly.h and add some white space. Signed-off-by: Grant Grundler <grundler@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:56:35 +04:00
#ifdef CONFIG_64BIT
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#define LDREG ldd
#define STREG std
#define LDREGX ldd,s
#define LDREGM ldd,mb
#define STREGM std,ma
#define SHRREG shrd
#define RP_OFFSET 16
#define FRAME_SIZE 128
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
#define CALLEE_REG_FRAME_SIZE 144
[PARISC] Specify level to fix binutils level promotion bug fixup.S needs to specify .level and use correct LDREG macro. New binutils has a bug where it doesn't "promote" from PA1.0 to PA1.1 correctly when using ",s" completer. remove use of __LP64__ in assembly.h and add some white space. Signed-off-by: Grant Grundler <grundler@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:56:35 +04:00
#else /* CONFIG_64BIT */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#define LDREG ldw
#define STREG stw
#define LDREGX ldwx,s
#define LDREGM ldwm
#define STREGM stwm
#define SHRREG shr
#define RP_OFFSET 20
#define FRAME_SIZE 64
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
#define CALLEE_REG_FRAME_SIZE 128
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#endif
[PARISC] Specify level to fix binutils level promotion bug fixup.S needs to specify .level and use correct LDREG macro. New binutils has a bug where it doesn't "promote" from PA1.0 to PA1.1 correctly when using ",s" completer. remove use of __LP64__ in assembly.h and add some white space. Signed-off-by: Grant Grundler <grundler@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:56:35 +04:00
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
#define CALLEE_SAVE_FRAME_SIZE (CALLEE_REG_FRAME_SIZE + CALLEE_FLOAT_FRAME_SIZE)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#ifdef CONFIG_PA20
[PARISC] Ensure all ldcw uses are ldcw,co on pa2.0 ldcw,co should always be used on pa2.0, otherwise the strict cache width alignment requirement is not relaxed. Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2006-04-22 10:48:22 +04:00
#define LDCW ldcw,co
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#define BL b,l
# ifdef CONFIG_64BIT
# define LEVEL 2.0w
# else
# define LEVEL 2.0
# endif
#else
[PARISC] Ensure all ldcw uses are ldcw,co on pa2.0 ldcw,co should always be used on pa2.0, otherwise the strict cache width alignment requirement is not relaxed. Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2006-04-22 10:48:22 +04:00
#define LDCW ldcw
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#define BL bl
#define LEVEL 1.1
#endif
#ifdef __ASSEMBLY__
#ifdef __LP64__
/* the 64-bit pa gnu assembler unfortunately defaults to .level 1.1 or 2.0 so
* work around that for now... */
.level 2.0w
#endif
kbuild: m68k,parisc,ppc,ppc64,s390,xtensa use generic asm-offsets.h support Delete obsoleted parts form arch makefiles and rename to asm-offsets.h Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2005-09-09 22:57:26 +04:00
#include <asm/asm-offsets.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#include <asm/page.h>
#include <asm/asmregs.h>
sp = 30
gp = 27
ipsw = 22
/*
* We provide two versions of each macro to convert from physical
* to virtual and vice versa. The "_r1" versions take one argument
* register, but trashes r1 to do the conversion. The other
* version takes two arguments: a src and destination register.
* However, the source and destination registers can not be
* the same register.
*/
.macro tophys grvirt, grphys
ldil L%(__PAGE_OFFSET), \grphys
sub \grvirt, \grphys, \grphys
.endm
.macro tovirt grphys, grvirt
ldil L%(__PAGE_OFFSET), \grvirt
add \grphys, \grvirt, \grvirt
.endm
.macro tophys_r1 gr
ldil L%(__PAGE_OFFSET), %r1
sub \gr, %r1, \gr
.endm
.macro tovirt_r1 gr
ldil L%(__PAGE_OFFSET), %r1
add \gr, %r1, \gr
.endm
.macro delay value
ldil L%\value, 1
ldo R%\value(1), 1
addib,UV,n -1,1,.
addib,NUV,n -1,1,.+8
nop
.endm
.macro debug value
.endm
/* Shift Left - note the r and t can NOT be the same! */
.macro shl r, sa, t
dep,z \r, 31-\sa, 32-\sa, \t
.endm
/* The PA 2.0 shift left */
.macro shlw r, sa, t
depw,z \r, 31-\sa, 32-\sa, \t
.endm
/* And the PA 2.0W shift left */
.macro shld r, sa, t
depd,z \r, 63-\sa, 64-\sa, \t
.endm
/* Shift Right - note the r and t can NOT be the same! */
.macro shr r, sa, t
extru \r, 31-\sa, 32-\sa, \t
.endm
/* pa20w version of shift right */
.macro shrd r, sa, t
extrd,u \r, 63-\sa, 64-\sa, \t
.endm
/* load 32-bit 'value' into 'reg' compensating for the ldil
* sign-extension when running in wide mode.
* WARNING!! neither 'value' nor 'reg' can be expressions
* containing '.'!!!! */
.macro load32 value, reg
ldil L%\value, \reg
ldo R%\value(\reg), \reg
.endm
.macro loadgp
#ifdef __LP64__
ldil L%__gp, %r27
ldo R%__gp(%r27), %r27
#else
ldil L%$global$, %r27
ldo R%$global$(%r27), %r27
#endif
.endm
#define SAVE_SP(r, where) mfsp r, %r1 ! STREG %r1, where
#define REST_SP(r, where) LDREG where, %r1 ! mtsp %r1, r
#define SAVE_CR(r, where) mfctl r, %r1 ! STREG %r1, where
#define REST_CR(r, where) LDREG where, %r1 ! mtctl %r1, r
.macro save_general regs
STREG %r1, PT_GR1 (\regs)
STREG %r2, PT_GR2 (\regs)
STREG %r3, PT_GR3 (\regs)
STREG %r4, PT_GR4 (\regs)
STREG %r5, PT_GR5 (\regs)
STREG %r6, PT_GR6 (\regs)
STREG %r7, PT_GR7 (\regs)
STREG %r8, PT_GR8 (\regs)
STREG %r9, PT_GR9 (\regs)
STREG %r10, PT_GR10(\regs)
STREG %r11, PT_GR11(\regs)
STREG %r12, PT_GR12(\regs)
STREG %r13, PT_GR13(\regs)
STREG %r14, PT_GR14(\regs)
STREG %r15, PT_GR15(\regs)
STREG %r16, PT_GR16(\regs)
STREG %r17, PT_GR17(\regs)
STREG %r18, PT_GR18(\regs)
STREG %r19, PT_GR19(\regs)
STREG %r20, PT_GR20(\regs)
STREG %r21, PT_GR21(\regs)
STREG %r22, PT_GR22(\regs)
STREG %r23, PT_GR23(\regs)
STREG %r24, PT_GR24(\regs)
STREG %r25, PT_GR25(\regs)
/* r26 is saved in get_stack and used to preserve a value across virt_map */
STREG %r27, PT_GR27(\regs)
STREG %r28, PT_GR28(\regs)
/* r29 is saved in get_stack and used to point to saved registers */
/* r30 stack pointer saved in get_stack */
STREG %r31, PT_GR31(\regs)
.endm
.macro rest_general regs
/* r1 used as a temp in rest_stack and is restored there */
LDREG PT_GR2 (\regs), %r2
LDREG PT_GR3 (\regs), %r3
LDREG PT_GR4 (\regs), %r4
LDREG PT_GR5 (\regs), %r5
LDREG PT_GR6 (\regs), %r6
LDREG PT_GR7 (\regs), %r7
LDREG PT_GR8 (\regs), %r8
LDREG PT_GR9 (\regs), %r9
LDREG PT_GR10(\regs), %r10
LDREG PT_GR11(\regs), %r11
LDREG PT_GR12(\regs), %r12
LDREG PT_GR13(\regs), %r13
LDREG PT_GR14(\regs), %r14
LDREG PT_GR15(\regs), %r15
LDREG PT_GR16(\regs), %r16
LDREG PT_GR17(\regs), %r17
LDREG PT_GR18(\regs), %r18
LDREG PT_GR19(\regs), %r19
LDREG PT_GR20(\regs), %r20
LDREG PT_GR21(\regs), %r21
LDREG PT_GR22(\regs), %r22
LDREG PT_GR23(\regs), %r23
LDREG PT_GR24(\regs), %r24
LDREG PT_GR25(\regs), %r25
LDREG PT_GR26(\regs), %r26
LDREG PT_GR27(\regs), %r27
LDREG PT_GR28(\regs), %r28
/* r29 points to register save area, and is restored in rest_stack */
/* r30 stack pointer restored in rest_stack */
LDREG PT_GR31(\regs), %r31
.endm
.macro save_fp regs
fstd,ma %fr0, 8(\regs)
fstd,ma %fr1, 8(\regs)
fstd,ma %fr2, 8(\regs)
fstd,ma %fr3, 8(\regs)
fstd,ma %fr4, 8(\regs)
fstd,ma %fr5, 8(\regs)
fstd,ma %fr6, 8(\regs)
fstd,ma %fr7, 8(\regs)
fstd,ma %fr8, 8(\regs)
fstd,ma %fr9, 8(\regs)
fstd,ma %fr10, 8(\regs)
fstd,ma %fr11, 8(\regs)
fstd,ma %fr12, 8(\regs)
fstd,ma %fr13, 8(\regs)
fstd,ma %fr14, 8(\regs)
fstd,ma %fr15, 8(\regs)
fstd,ma %fr16, 8(\regs)
fstd,ma %fr17, 8(\regs)
fstd,ma %fr18, 8(\regs)
fstd,ma %fr19, 8(\regs)
fstd,ma %fr20, 8(\regs)
fstd,ma %fr21, 8(\regs)
fstd,ma %fr22, 8(\regs)
fstd,ma %fr23, 8(\regs)
fstd,ma %fr24, 8(\regs)
fstd,ma %fr25, 8(\regs)
fstd,ma %fr26, 8(\regs)
fstd,ma %fr27, 8(\regs)
fstd,ma %fr28, 8(\regs)
fstd,ma %fr29, 8(\regs)
fstd,ma %fr30, 8(\regs)
fstd %fr31, 0(\regs)
.endm
.macro rest_fp regs
fldd 0(\regs), %fr31
fldd,mb -8(\regs), %fr30
fldd,mb -8(\regs), %fr29
fldd,mb -8(\regs), %fr28
fldd,mb -8(\regs), %fr27
fldd,mb -8(\regs), %fr26
fldd,mb -8(\regs), %fr25
fldd,mb -8(\regs), %fr24
fldd,mb -8(\regs), %fr23
fldd,mb -8(\regs), %fr22
fldd,mb -8(\regs), %fr21
fldd,mb -8(\regs), %fr20
fldd,mb -8(\regs), %fr19
fldd,mb -8(\regs), %fr18
fldd,mb -8(\regs), %fr17
fldd,mb -8(\regs), %fr16
fldd,mb -8(\regs), %fr15
fldd,mb -8(\regs), %fr14
fldd,mb -8(\regs), %fr13
fldd,mb -8(\regs), %fr12
fldd,mb -8(\regs), %fr11
fldd,mb -8(\regs), %fr10
fldd,mb -8(\regs), %fr9
fldd,mb -8(\regs), %fr8
fldd,mb -8(\regs), %fr7
fldd,mb -8(\regs), %fr6
fldd,mb -8(\regs), %fr5
fldd,mb -8(\regs), %fr4
fldd,mb -8(\regs), %fr3
fldd,mb -8(\regs), %fr2
fldd,mb -8(\regs), %fr1
fldd,mb -8(\regs), %fr0
.endm
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
.macro callee_save_float
fstd,ma %fr12, 8(%r30)
fstd,ma %fr13, 8(%r30)
fstd,ma %fr14, 8(%r30)
fstd,ma %fr15, 8(%r30)
fstd,ma %fr16, 8(%r30)
fstd,ma %fr17, 8(%r30)
fstd,ma %fr18, 8(%r30)
fstd,ma %fr19, 8(%r30)
fstd,ma %fr20, 8(%r30)
fstd,ma %fr21, 8(%r30)
.endm
.macro callee_rest_float
fldd,mb -8(%r30), %fr21
fldd,mb -8(%r30), %fr20
fldd,mb -8(%r30), %fr19
fldd,mb -8(%r30), %fr18
fldd,mb -8(%r30), %fr17
fldd,mb -8(%r30), %fr16
fldd,mb -8(%r30), %fr15
fldd,mb -8(%r30), %fr14
fldd,mb -8(%r30), %fr13
fldd,mb -8(%r30), %fr12
.endm
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#ifdef __LP64__
.macro callee_save
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
std,ma %r3, CALLEE_REG_FRAME_SIZE(%r30)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
mfctl %cr27, %r3
std %r4, -136(%r30)
std %r5, -128(%r30)
std %r6, -120(%r30)
std %r7, -112(%r30)
std %r8, -104(%r30)
std %r9, -96(%r30)
std %r10, -88(%r30)
std %r11, -80(%r30)
std %r12, -72(%r30)
std %r13, -64(%r30)
std %r14, -56(%r30)
std %r15, -48(%r30)
std %r16, -40(%r30)
std %r17, -32(%r30)
std %r18, -24(%r30)
std %r3, -16(%r30)
.endm
.macro callee_rest
ldd -16(%r30), %r3
ldd -24(%r30), %r18
ldd -32(%r30), %r17
ldd -40(%r30), %r16
ldd -48(%r30), %r15
ldd -56(%r30), %r14
ldd -64(%r30), %r13
ldd -72(%r30), %r12
ldd -80(%r30), %r11
ldd -88(%r30), %r10
ldd -96(%r30), %r9
ldd -104(%r30), %r8
ldd -112(%r30), %r7
ldd -120(%r30), %r6
ldd -128(%r30), %r5
ldd -136(%r30), %r4
mtctl %r3, %cr27
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
ldd,mb -CALLEE_REG_FRAME_SIZE(%r30), %r3
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
.endm
#else /* ! __LP64__ */
.macro callee_save
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
stw,ma %r3, CALLEE_REG_FRAME_SIZE(%r30)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
mfctl %cr27, %r3
stw %r4, -124(%r30)
stw %r5, -120(%r30)
stw %r6, -116(%r30)
stw %r7, -112(%r30)
stw %r8, -108(%r30)
stw %r9, -104(%r30)
stw %r10, -100(%r30)
stw %r11, -96(%r30)
stw %r12, -92(%r30)
stw %r13, -88(%r30)
stw %r14, -84(%r30)
stw %r15, -80(%r30)
stw %r16, -76(%r30)
stw %r17, -72(%r30)
stw %r18, -68(%r30)
stw %r3, -64(%r30)
.endm
.macro callee_rest
ldw -64(%r30), %r3
ldw -68(%r30), %r18
ldw -72(%r30), %r17
ldw -76(%r30), %r16
ldw -80(%r30), %r15
ldw -84(%r30), %r14
ldw -88(%r30), %r13
ldw -92(%r30), %r12
ldw -96(%r30), %r11
ldw -100(%r30), %r10
ldw -104(%r30), %r9
ldw -108(%r30), %r8
ldw -112(%r30), %r7
ldw -116(%r30), %r6
ldw -120(%r30), %r5
ldw -124(%r30), %r4
mtctl %r3, %cr27
[PARISC] Fix the alloc_slabmgmt panic Fix the alloc_slabmgmt panic Hopefully this should also fix a lot of other intermittent kernel bugs. The problem has been around since 2.6.9-rc2-pa6 when we allowed floating point registers to be used in kernel code. The essence of the problem is that gcc prefers to use floating point for integer divides and multiples. Further, it can rely on the values in the no clobber fp regs being correct across a function call. Unfortunately, our task switch function only saves the integer no clobber registers, not the fp ones, so if gcc makes a function call to any function in the kernel which could sleep, the values it is relying on in any no clobber floating point register may be lost. In the case of alloc_slabmgmt, the value of the page offset is being stored in %fr12 across a call to kmem_getpages(), which sleeps if no pages are available. Thus, the offset can be trashed and the slab code can end up with a completely bogus address leading to corruption. Kudos to Randolph who came up with the program to trip this problem at will and thus allowed it to be tracked and fixed. Signed-off-by: James Bottomley <jejb@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:53:26 +04:00
ldw,mb -CALLEE_REG_FRAME_SIZE(%r30), %r3
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
.endm
#endif /* ! __LP64__ */
.macro save_specials regs
SAVE_SP (%sr0, PT_SR0 (\regs))
SAVE_SP (%sr1, PT_SR1 (\regs))
SAVE_SP (%sr2, PT_SR2 (\regs))
SAVE_SP (%sr3, PT_SR3 (\regs))
SAVE_SP (%sr4, PT_SR4 (\regs))
SAVE_SP (%sr5, PT_SR5 (\regs))
SAVE_SP (%sr6, PT_SR6 (\regs))
SAVE_SP (%sr7, PT_SR7 (\regs))
SAVE_CR (%cr17, PT_IASQ0(\regs))
mtctl %r0, %cr17
SAVE_CR (%cr17, PT_IASQ1(\regs))
SAVE_CR (%cr18, PT_IAOQ0(\regs))
mtctl %r0, %cr18
SAVE_CR (%cr18, PT_IAOQ1(\regs))
#ifdef __LP64__
/* cr11 (sar) is a funny one. 5 bits on PA1.1 and 6 bit on PA2.0
* For PA2.0 mtsar or mtctl always write 6 bits, but mfctl only
* reads 5 bits. Use mfctl,w to read all six bits. Otherwise
* we lose the 6th bit on a save/restore over interrupt.
*/
mfctl,w %cr11, %r1
STREG %r1, PT_SAR (\regs)
#else
SAVE_CR (%cr11, PT_SAR (\regs))
#endif
SAVE_CR (%cr19, PT_IIR (\regs))
/*
* Code immediately following this macro (in intr_save) relies
* on r8 containing ipsw.
*/
mfctl %cr22, %r8
STREG %r8, PT_PSW(\regs)
.endm
.macro rest_specials regs
REST_SP (%sr0, PT_SR0 (\regs))
REST_SP (%sr1, PT_SR1 (\regs))
REST_SP (%sr2, PT_SR2 (\regs))
REST_SP (%sr3, PT_SR3 (\regs))
REST_SP (%sr4, PT_SR4 (\regs))
REST_SP (%sr5, PT_SR5 (\regs))
REST_SP (%sr6, PT_SR6 (\regs))
REST_SP (%sr7, PT_SR7 (\regs))
REST_CR (%cr17, PT_IASQ0(\regs))
REST_CR (%cr17, PT_IASQ1(\regs))
REST_CR (%cr18, PT_IAOQ0(\regs))
REST_CR (%cr18, PT_IAOQ1(\regs))
REST_CR (%cr11, PT_SAR (\regs))
REST_CR (%cr22, PT_PSW (\regs))
.endm
[PARISC] Make sure use of RFI conforms to PA 2.0 and 1.1 arch docs 2.6.12-rc4-pa3 : first pass at making sure use of RFI conforms to PA 2.0 arch pages F-4 and F-5, PA 1.1 Arch page 3-19 and 3-20. The discussion revolves around all the rules for clearing PSW Q-bit. The hard part is meeting all the rules for "relied upon translation". .align directive is used to guarantee the critical sequence ends more than 8 instructions (32 bytes) from the end of page. Signed-off-by: Grant Grundler <grundler@parisc-linux.org> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2005-10-22 06:40:07 +04:00
/* First step to create a "relied upon translation"
* See PA 2.0 Arch. page F-4 and F-5.
*
* The ssm was originally necessary due to a "PCxT bug".
* But someone decided it needed to be added to the architecture
* and this "feature" went into rev3 of PA-RISC 1.1 Arch Manual.
* It's been carried forward into PA 2.0 Arch as well. :^(
*
* "ssm 0,%r0" is a NOP with side effects (prefetch barrier).
* rsm/ssm prevents the ifetch unit from speculatively fetching
* instructions past this line in the code stream.
* PA 2.0 processor will single step all insn in the same QUAD (4 insn).
*/
.macro pcxt_ssm_bug
rsm PSW_SM_I,%r0
nop /* 1 */
nop /* 2 */
nop /* 3 */
nop /* 4 */
nop /* 5 */
nop /* 6 */
nop /* 7 */
.endm
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 02:20:36 +04:00
#endif /* __ASSEMBLY__ */
#endif
Reference in New Issue Copy Permalink