Thread (3 messages) 3 messages, 2 authors, 2019-09-30

Re: Regression: Network link not coming up after suspend/resume cycle

From: Jan Janssen <hidden>
Date: 2019-09-30 21:18:59

On Monday, 30 September 2019 01:26:21 CEST Vladimir Oltean wrote:
Hi Jan,

On Sun, 29 Sep 2019 at 22:25, Jan Janssen [off-list ref] wrote:
quoted
Hi,

I've been noticing lately that my network link sometimes does not go up
after a suspend resume cycle (roughly 1 or 2 out of 10 times). This also
sometimes happens with a fresh boot too. Doing a manual
"ip link set down/up" cycle resolves this issue.

I was able to bisect it to the commit below (or hope so) and also CCed
the maintainer for my driver too.

This is happening on a up-to-date Arch Linux system with a Intel I219-V.

Jan



7ede7b03484bbb035aa5be98c45a40cfabdc0738 is the first bad commit
commit 7ede7b03484bbb035aa5be98c45a40cfabdc0738
Author: Vedang Patel [off-list ref]
Date:   Tue Jun 25 15:07:18 2019 -0700

taprio: make clock reference conversions easier

Later in this series we will need to transform from
CLOCK_MONOTONIC (used in TCP) to the clock reference used in TAPRIO.

Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Vedang Patel <redacted>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/sched/sch_taprio.c | 30 ++++++++++++++++++++++--------
1 file changed, 22 insertions(+), 8 deletions(-)
That is a mechanical patch that produces no behavior change.
Furthermore, even if distributions were to build with
CONFIG_NET_SCH_TAPRIO (which there aren't many reasons to), it is
extremely likely that this qdisc is not enabled by default on your
interface. Are you voluntarily using taprio?
You might need to bisect again.

Regards,
-Vladimir
You were right. The module wasn't even built.

I did a more careful bisection this time and got me this commit:

59653e6497d16f7ac1d9db088f3959f57ee8c3db is the first bad commit
commit 59653e6497d16f7ac1d9db088f3959f57ee8c3db
Author: Detlev Casanova [off-list ref]
Date:   Sat Jun 22 23:14:37 2019 -0400

    e1000e: Make watchdog use delayed work

    Use delayed work instead of timers to run the watchdog of the e1000e
    driver.

    Simplify the code with one less middle function.

    Signed-off-by: Detlev Casanova [off-list ref]
    Tested-by: Aaron Brown [off-list ref]
    Signed-off-by: Jeff Kirsher [off-list ref]

 drivers/net/ethernet/intel/e1000e/e1000.h  |  5 +--
 drivers/net/ethernet/intel/e1000e/netdev.c | 54 +++++++++++++++
+--------------
 2 files changed, 32 insertions(+), 27 deletions(-)


This time I also went and reverted it on top of v5.3, which successfully made
the issue go away. I've CC'ed the author of the patch.

Jan

Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help