linux-kernel - Re: Linux v2.6.18-rc5

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

Date:	Sun, 03 Sep 2006 15:10:44 +0900
From:	Tejun Heo <htejun@...il.com>
To:	Keith Owens <kaos@....com.au>
CC:	Andrew Morton <akpm@...l.org>, Linus Torvalds <torvalds@...l.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Jeff Garzik <jeff@...zik.org>
Subject: Re: Linux v2.6.18-rc5

Hello, Keith Owens.

Sorry about late respond.  I'm quite occupied with personal stuff these 
days.

Keith Owens wrote:
>>> (2) I have seen the same intermittent bug on ICH7 SATA but
>>>     PIIX_FLAG_IGNORE_PCS is only set for ich5 and i6300esb_sata.  It
>>>     probably needs to be set for ich7 as well.
>> No, ICH7 up to this point has been believed to have well-behaving PCS. 
>> If you report PCS problem, you'll be the first.  Also, note that ICH7 
>> suffers from ghost device probing problem if PCS is not honored exactly. 
>>  Are you sure it's the same problem?
> 
> It definitely looks like it.  Stock 2.6.18-rc5 plus this patch to
> activate ata_debug from boot until just after probing drives.
[--snip--]
> <3>piix_sata_present_mask: ata1: ENTER, pcs=0x15 base=0
> <3>piix_sata_present_mask: ata1: LEAVE, pcs=0x15 present_mask=0x3
> <3>piix_sata_present_mask: ata1: ENTER, pcs=0x0 base=0
> <3>piix_sata_present_mask: ata1: LEAVE, pcs=0x0 present_mask=0x3
> <3>piix_sata_present_mask: ata1: ENTER, pcs=0x15 base=0
> <3>piix_sata_present_mask: ata1: LEAVE, pcs=0x15 present_mask=0x3
> <3>piix_sata_present_mask: ata1: ENTER, pcs=0x0 base=0
> <3>piix_sata_present_mask: ata1: LEAVE, pcs=0x0 present_mask=0x3
> <3>piix_sata_present_mask: ata1: ENTER, pcs=0x15 base=0

Yeah, it definitely looks like it.  Does the kernel print a message 
which looks like the following before those debug messages?

"updating PCS from 0x0 to 0x5"

> Note the pcs=0x0 values.  Adding PIIX_FLAG_IGNORE_PCS to
> ich6m_sata_ahci gets past the failure to detect pcs, with no sign of
> any ghost devices.  BTW, dropping down to 2.6.17 with the same config
> has no problem detecting the disk, even without PIIX_FLAG_IGNORE_PCS on
> ich6m_sata_ahci.

BTW, you can set PIIX_FLAG_IGNORE_PCS by using force_pcs=1 module 
parameter now.

> lspci extract, this is an ICH7M.
> 
> 00:1f.0 Class 0601: 8086:27b9 (rev 02)
> 	Subsystem: 1033:832c
> 	Flags: bus master, medium devsel, latency 0
> 	Capabilities: [e0] Vendor Specific Information
> 
> 00:1f.2 Class 0101: 8086:27c4 (rev 02) (prog-if 80)
> 	Subsystem: 1033:832c
> 	Flags: bus master, 66MHz, medium devsel, latency 0, IRQ 18
> 	I/O ports at <unassigned>
> 	I/O ports at <unassigned>
> 	I/O ports at <unassigned>
> 	I/O ports at <unassigned>
> 	I/O ports at 18b0 [size=16]
> 	Capabilities: [70] Power Management version 2
> 
> 00:1f.3 Class 0c05: 8086:27da (rev 02)
> 	Subsystem: 1033:832c
> 	Flags: medium devsel, IRQ 11
> 	I/O ports at 18c0 [size=32]
> 

Hmm... Can you try the attached patch and see what happens?  ATM, I'm on 
the road and can't test the patch, so it's only compile-tested.  This 
patch basically reverts some of the effects of the following commit and 
makes PCS update a little bit more aggressive iff necessary.

ea35d29e2fa8b3d766a2ce8fbcce599dce8d2734
[libata] ata_piix: Consolidate PCS register writing

If this works for you ich7m, can you please test this on your formerly 
problematic ich5 with force_pcs=2 specified?  I initially thought that 
the ich5 problem was caused by exact PCS map change and thus added 
IGNORE_PCS as workaround but if the same problem occurs on ich7 and is 
fixed by the attached patch, it's due to conservative PCS update change 
and thus the original IGNORE_PCS fix on ich5 might not be necessary.

Sorry this PCS business causes you and other people so much trouble.  It 
just doesn't work quite as ata_piix developers expect.

Thanks.

-- 
tejun

View attachment "patch" of type "text/plain" (3418 bytes)