Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.kernel > #67059

Bug#422675: Possible fix by reverting a patch that entered mainline between 3.11 and 3.12

Cross-posted to 2 groups.

Show all headers | View raw

[Multipart message — attachments visible in raw view] - view raw

Control: reassign -1 src:linux

On Lu, 11 mai 20, 18:03:00, Jim Knopf wrote:
> ­In case anyone stumbles upon this thread looking for a solution,
> there has been success here reverting a particular kernel patch
> to sis900.c that occured between linux versions 3.11 and 3.12.
> 
> The machine tested has been a Gericom Webboy (laptop from
> year ~2000), it was connected to a tplink router.
> 
> Systems with kernel versions 2.6.21.7, 2.6.27.27, 3.8.2 and 3.11
> had acceptable network performance, while a regression was
> observed after booting 4.4.0 vanilla.  With this version the net-
> work adapter delivered a mere 100 kb/s, while in the megabyte
> range with aforementioned versions.
> 
> Backtracking the issue down to kernel 3.12.6 reliably exhibited
> the same degradation for each kernel tested - flooding dmesg
> log with the log patterns described earlier in this bug thread.
> 
> Visiting the kernel changelogs between versions 3.11 and 3.12.6
> the following candidate patch could be identified as a possible
> cause:
> https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/drivers/net/ethernet/sis/sis900.c?h=v3.12&id=78f1ccc4f82cd7bbb65b0e938ed0dfd81e902fe0
> 
> The patch title is
> sis900: don't restart auto-negotiation each time after link resume.
> 
> .. applied to linux mainline on
> 2013-08-20
> 
> Once this patch is reverted and 3.12.6 kernel recompiled, the
> sis900 network card's performance was restored.  This observation
> has then been confirmed and reproduced by reverse applying the
> patch to kernel 4.4.0.
> 
> This patch may or may not be the cause for other setups employing
> sis900 as a network chip, but has been reproducibly identified as
> cause for performance degradation for the issue above.
> 
> 
> The mii-tool workaround described earlier has worked as well, i.e.
> without patching the sis900.c driver, setting 10baseT-FD only for
> auto-negotiation events.  (100baseTx-HD showed no improvement).
> 10baseT-FD limits the network card's speed to about one mbyte/s,
> which is inferior to proper patching of sis900 kernel driver.
> 
> 
> Regards,
> Jim

-- 
Looking after bugs filled against unknown packages

Back to linux.debian.kernel | Previous | Next — Next in thread | Find similar | Unroll thread

Thread

Bug#422675: Possible fix by reverting a patch that entered mainline between 3.11 and 3.12 Andrei POPESCU <andreimpopescu@gmail.com> - 2020-05-11 18:50 +0200
  Processed: Re: Bug#422675: Possible fix by reverting a patch that  entered mainline between 3.11 and 3.12 "Debian Bug Tracking System" <owner@bugs.debian.org> - 2020-05-11 18:50 +0200

csiph-web