Netdev List
 help / color / mirror / Atom feed
* [PATCH] net: revert default NAPI poll timeout to 2 jiffies
@ 2020-04-06 11:39 Konstantin Khlebnikov
  2020-04-08  1:31 ` David Miller
  0 siblings, 1 reply; 2+ messages in thread
From: Konstantin Khlebnikov @ 2020-04-06 11:39 UTC (permalink / raw)
  To: netdev, linux-kernel
  Cc: Jakub Kicinski, Dmitry Yakunin, David S. Miller,
	Matthew Whitehead

For HZ < 1000 timeout 2000us rounds up to 1 jiffy but expires randomly
because next timer interrupt could come shortly after starting softirq.

For commonly used CONFIG_HZ=1000 nothing changes.

Fixes: 7acf8a1e8a28 ("Replace 2 jiffies with sysctl netdev_budget_usecs to enable softirq tuning")
Reported-by: Dmitry Yakunin <zeil@yandex-team.ru>
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
---
 net/core/dev.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index 9c9e763bfe0e..df8097b8e286 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4140,7 +4140,8 @@ EXPORT_SYMBOL(netdev_max_backlog);
 
 int netdev_tstamp_prequeue __read_mostly = 1;
 int netdev_budget __read_mostly = 300;
-unsigned int __read_mostly netdev_budget_usecs = 2000;
+/* Must be at least 2 jiffes to guarantee 1 jiffy timeout */
+unsigned int __read_mostly netdev_budget_usecs = 2 * USEC_PER_SEC / HZ;
 int weight_p __read_mostly = 64;           /* old backlog weight */
 int dev_weight_rx_bias __read_mostly = 1;  /* bias for backlog weight */
 int dev_weight_tx_bias __read_mostly = 1;  /* bias for output_queue quota */


^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2020-04-08  1:31 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-04-06 11:39 [PATCH] net: revert default NAPI poll timeout to 2 jiffies Konstantin Khlebnikov
2020-04-08  1:31 ` David Miller

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox