gve: Account for stopped queues when reading NIC stats

We now account for the fact that the NIC might send us stats for a
subset of queues. Without this change, gve_get_ethtool_stats might make
an invalid access on the priv->stats_report->stats array.

Tested-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Signed-off-by: Shailend Chand <shailend@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
Shailend Chand 2024-05-01 23:25:47 +00:00 committed by David S. Miller
parent 770f52d5a0
commit af9bcf910b

View File

@ -8,6 +8,7 @@
#include "gve.h"
#include "gve_adminq.h"
#include "gve_dqo.h"
#include "gve_utils.h"
static void gve_get_drvinfo(struct net_device *netdev,
struct ethtool_drvinfo *info)
@ -165,6 +166,8 @@ gve_get_ethtool_stats(struct net_device *netdev,
struct stats *report_stats;
int *rx_qid_to_stats_idx;
int *tx_qid_to_stats_idx;
int num_stopped_rxqs = 0;
int num_stopped_txqs = 0;
struct gve_priv *priv;
bool skip_nic_stats;
unsigned int start;
@ -181,12 +184,23 @@ gve_get_ethtool_stats(struct net_device *netdev,
sizeof(int), GFP_KERNEL);
if (!rx_qid_to_stats_idx)
return;
for (ring = 0; ring < priv->rx_cfg.num_queues; ring++) {
rx_qid_to_stats_idx[ring] = -1;
if (!gve_rx_was_added_to_block(priv, ring))
num_stopped_rxqs++;
}
tx_qid_to_stats_idx = kmalloc_array(num_tx_queues,
sizeof(int), GFP_KERNEL);
if (!tx_qid_to_stats_idx) {
kfree(rx_qid_to_stats_idx);
return;
}
for (ring = 0; ring < num_tx_queues; ring++) {
tx_qid_to_stats_idx[ring] = -1;
if (!gve_tx_was_added_to_block(priv, ring))
num_stopped_txqs++;
}
for (rx_pkts = 0, rx_bytes = 0, rx_hsplit_pkt = 0,
rx_skb_alloc_fail = 0, rx_buf_alloc_fail = 0,
rx_desc_err_dropped_pkt = 0, rx_hsplit_unsplit_pkt = 0,
@ -260,7 +274,13 @@ gve_get_ethtool_stats(struct net_device *netdev,
/* For rx cross-reporting stats, start from nic rx stats in report */
base_stats_idx = GVE_TX_STATS_REPORT_NUM * num_tx_queues +
GVE_RX_STATS_REPORT_NUM * priv->rx_cfg.num_queues;
max_stats_idx = NIC_RX_STATS_REPORT_NUM * priv->rx_cfg.num_queues +
/* The boundary between driver stats and NIC stats shifts if there are
* stopped queues.
*/
base_stats_idx += NIC_RX_STATS_REPORT_NUM * num_stopped_rxqs +
NIC_TX_STATS_REPORT_NUM * num_stopped_txqs;
max_stats_idx = NIC_RX_STATS_REPORT_NUM *
(priv->rx_cfg.num_queues - num_stopped_rxqs) +
base_stats_idx;
/* Preprocess the stats report for rx, map queue id to start index */
skip_nic_stats = false;
@ -274,6 +294,10 @@ gve_get_ethtool_stats(struct net_device *netdev,
skip_nic_stats = true;
break;
}
if (queue_id < 0 || queue_id >= priv->rx_cfg.num_queues) {
net_err_ratelimited("Invalid rxq id in NIC stats\n");
continue;
}
rx_qid_to_stats_idx[queue_id] = stats_idx;
}
/* walk RX rings */
@ -308,11 +332,11 @@ gve_get_ethtool_stats(struct net_device *netdev,
data[i++] = rx->rx_copybreak_pkt;
data[i++] = rx->rx_copied_pkt;
/* stats from NIC */
if (skip_nic_stats) {
stats_idx = rx_qid_to_stats_idx[ring];
if (skip_nic_stats || stats_idx < 0) {
/* skip NIC rx stats */
i += NIC_RX_STATS_REPORT_NUM;
} else {
stats_idx = rx_qid_to_stats_idx[ring];
for (j = 0; j < NIC_RX_STATS_REPORT_NUM; j++) {
u64 value =
be64_to_cpu(report_stats[stats_idx + j].value);
@ -338,7 +362,8 @@ gve_get_ethtool_stats(struct net_device *netdev,
/* For tx cross-reporting stats, start from nic tx stats in report */
base_stats_idx = max_stats_idx;
max_stats_idx = NIC_TX_STATS_REPORT_NUM * num_tx_queues +
max_stats_idx = NIC_TX_STATS_REPORT_NUM *
(num_tx_queues - num_stopped_txqs) +
max_stats_idx;
/* Preprocess the stats report for tx, map queue id to start index */
skip_nic_stats = false;
@ -352,6 +377,10 @@ gve_get_ethtool_stats(struct net_device *netdev,
skip_nic_stats = true;
break;
}
if (queue_id < 0 || queue_id >= num_tx_queues) {
net_err_ratelimited("Invalid txq id in NIC stats\n");
continue;
}
tx_qid_to_stats_idx[queue_id] = stats_idx;
}
/* walk TX rings */
@ -383,11 +412,11 @@ gve_get_ethtool_stats(struct net_device *netdev,
data[i++] = gve_tx_load_event_counter(priv, tx);
data[i++] = tx->dma_mapping_error;
/* stats from NIC */
if (skip_nic_stats) {
stats_idx = tx_qid_to_stats_idx[ring];
if (skip_nic_stats || stats_idx < 0) {
/* skip NIC tx stats */
i += NIC_TX_STATS_REPORT_NUM;
} else {
stats_idx = tx_qid_to_stats_idx[ring];
for (j = 0; j < NIC_TX_STATS_REPORT_NUM; j++) {
u64 value =
be64_to_cpu(report_stats[stats_idx + j].value);