linux-kernel - s2ram still broken with CONFIG_NO

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1173645941.4204.5.camel@localhost>
Date:	Sun, 11 Mar 2007 21:45:41 +0100
From:	Soeren Sonnenburg <kernel@....de>
To:	Linux Kernel <linux-kernel@...r.kernel.org>
Cc:	Thomas Gleixner <tglx@...utronix.de>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Adrian Bunk <bunk@...sta.de>,
	Michal Piotrowski <michal.k.k.piotrowski@...il.com>,
	Emil Karlson <jkarlson@...hut.fi>,
	"Michael S\. Tsirkin" <mst@...lanox.co.il>,
	Len Brown <lenb@...nel.org>, Tejun Heo <htejun@...il.com>
Subject: s2ram still broken with CONFIG_NO_HZ / HPET (macbook pro)

(sorry if I am CC'ing the wrong people, seemed for me this is related to
rc-2 regression 5/6)

After a very long compile sessions (is there something like git bisect
to speed this up?) to find the config options that break s2ram between
my minimalistic and full config I finally have some results:

As soon as I enable

CONFIG_TICK_ONESHOT=y
CONFIG_NO_HZ=y

s2ram does not anymore return from resume (just hangs).

happens with git-current (i.e last commit is
commit be521466feb3bb1cd89de82a2b1d080e9ebd3cb6
Author: Linus Torvalds <torvalds@...dy.linux-foundation.org>
Date:   Sat Mar 10 14:22:07 2007 -0800)

I am attaching a diff between the working and broken config and the
working config itself.

As I suspected also something broken HPET wise, I enabled 

CONFIG_TICK_ONESHOT=y
CONFIG_HIGH_RES_TIMERS=y
CONFIG_HPET_TIMER=y
CONFIG_HPET_EMULATE_RTC=y
CONFIG_HPET=y
CONFIG_HPET_MMAP=y

( admittingly these two option too,
CONFIG_X86_MCE=y
CONFIG_X86_MCE_P4THERMAL=y )

and now see the same hang (config diffs attached)

I then disabled the HPET stuff again and voila the kernel resumes from
s2ram even with *MCE=y.

Note that this can be reproduced with a minimalistic config... anyway as
the intersection is CONFIG_TICK_ONESHOT I would be hoping that this is
already narrowed down enough...

I am not sure whether this is related, but when I boot with HPET stuff
enabled it is *sometimes* necessary to put noapic as the kernel options
line (else I see hangs asking me to rerun things with apic=debug)

Elsewise I still see the

ATA: abnormal status 0x80 on port 0x000140df
ATA: abnormal status 0x80 on port 0x000140df
ata1.00: configured for UDMA/33
ata3.01: revalidation failed (errno=-2)
ata3: failed to recover some devices, retrying in 5 secs
ATA: abnormal status 0x7F on port 0x000140df
ATA: abnormal status 0x7F on port 0x000140df
ata3.01: configured for UDMA/133

messages (making me shiver) directly after resume, the system is
reliably functioning though...

Soeren
-- 
Sometimes, there's a moment as you're waking, when you become aware of
the real world around you, but you're still dreaming.

View attachment "NO_HZ_breaks_s2ram.txt" of type "text/x-patch" (574 bytes)

View attachment "working-config" of type "text/plain" (52860 bytes)

View attachment "HPET_breaks_s2ram.txt" of type "text/x-patch" (1419 bytes)