github-mirror/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-07-28 18:21:24 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity Graph

net: napi: Avoid gro timer misfiring at end of busypoll

Browse Source 
When in irq deferral mode (defer-hard-irqs > 0), a short enough
gro-flush timeout can trigger before NAPI_STATE_SCHED is cleared if the
last poll in busy_poll_stop() takes too long. This can have the effect
of leaving the queue stuck with interrupts disabled and no timer armed
which results in a tx timeout if there is no subsequent busypoll cycle.

To prevent this, defer the gro-flush timer arm after the last poll.

Fixes: 7fd3253a7d ("net: Introduce preferred busy-polling")
Co-developed-by: Martin Karsten <mkarsten@uwaterloo.ca>
Signed-off-by: Martin Karsten <mkarsten@uwaterloo.ca>
Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
Reviewed-by: Joe Damato <joe@dama.to>
Link: https://patch.msgid.link/20260506090808.820559-2-dtatulea@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This commit is contained in:
Dragos Tatulea 2026-05-06 09:08:08 +00:00 committed by Jakub Kicinski
parent dffddaa0ce
commit 58e2330bd4
1 changed files with 12 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
net/core/dev.c
View File

@ -6862,9 +6862,9 @@ static void skb_defer_free_flush(void)
#if defined(CONFIG_NET_RX_BUSY_POLL) #if defined(CONFIG_NET_RX_BUSY_POLL)
static void __busy_poll_stop(struct napi_struct *napi, bool skip_schedule) static void __busy_poll_stop(struct napi_struct *napi, unsigned long timeout)
{ {
if (!skip_schedule) { if (!timeout) {
gro_normal_list(&napi->gro); gro_normal_list(&napi->gro);
__napi_schedule(napi); __napi_schedule(napi);
return; return;
@ -6874,6 +6874,8 @@ static void __busy_poll_stop(struct napi_struct *napi, bool skip_schedule)
gro_flush_normal(&napi->gro, HZ >= 1000); gro_flush_normal(&napi->gro, HZ >= 1000);
clear_bit(NAPI_STATE_SCHED, &napi->state); clear_bit(NAPI_STATE_SCHED, &napi->state);
hrtimer_start(&napi->timer, ns_to_ktime(timeout),
HRTIMER_MODE_REL_PINNED);
} }
enum { enum {
@ -6885,8 +6887,7 @@ static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock,
unsigned flags, u16 budget) unsigned flags, u16 budget)
{ {
struct bpf_net_context __bpf_net_ctx, *bpf_net_ctx; struct bpf_net_context __bpf_net_ctx, *bpf_net_ctx;
bool skip_schedule = false; unsigned long timeout = 0;
unsigned long timeout;
int rc; int rc;
/* Busy polling means there is a high chance device driver hard irq /* Busy polling means there is a high chance device driver hard irq
@ -6906,10 +6907,12 @@ static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock,
if (flags & NAPI_F_PREFER_BUSY_POLL) { if (flags & NAPI_F_PREFER_BUSY_POLL) {
napi->defer_hard_irqs_count = napi_get_defer_hard_irqs(napi); napi->defer_hard_irqs_count = napi_get_defer_hard_irqs(napi);
timeout = napi_get_gro_flush_timeout(napi); if (napi->defer_hard_irqs_count) {
if (napi->defer_hard_irqs_count && timeout) { /* A short enough gro flush timeout and long enough
hrtimer_start(&napi->timer, ns_to_ktime(timeout), HRTIMER_MODE_REL_PINNED); * poll can result in timer firing too early.
skip_schedule = true; * Timer will be armed later if necessary.
*/
timeout = napi_get_gro_flush_timeout(napi);
} }
} }
@ -6924,7 +6927,7 @@ static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock,
trace_napi_poll(napi, rc, budget); trace_napi_poll(napi, rc, budget);
netpoll_poll_unlock(have_poll_lock); netpoll_poll_unlock(have_poll_lock);
if (rc == budget) if (rc == budget)
__busy_poll_stop(napi, skip_schedule); __busy_poll_stop(napi, timeout);
bpf_net_ctx_clear(bpf_net_ctx); bpf_net_ctx_clear(bpf_net_ctx);
local_bh_enable(); local_bh_enable();
} }