linux-kernel - [bootup crash, -git] Re: patch pci-pcie-aspm-support.patch added to gregkh-2.6 tree

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 for Android: free password hash cracker in your pocket

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

Date:	Sat, 2 Feb 2008 11:55:06 +0100
From:	Ingo Molnar <mingo@...e.hu>
To:	gregkh@...e.de
Cc:	shaohua.li@...el.com, auke-jan.h.kok@...el.com, greg@...ah.com,
	linux-kernel@...r.kernel.org, linux-pci@...ey.karlin.mff.cuni.cz,
	matthew@....cx, venkatesh.pallipadi@...el.com,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Andrew Morton <akpm@...ux-foundation.org>
Subject: [bootup crash, -git] Re: patch pci-pcie-aspm-support.patch added
	to gregkh-2.6 tree


* gregkh@...e.de <gregkh@...e.de> wrote:

> This is a note to let you know that I've just added the patch titled
> 
>      Subject: PCI: PCIE ASPM support
> 
> to my gregkh-2.6 tree.  Its filename is
> 
>      pci-pcie-aspm-support.patch

uhm. One week ago this patch was added to your PCI tree. It never 
touched -mm AFAICS and today it was merged upstream (commit 
6c723d5bd89f03fc3ef627d50f89ade054d2ee3b):

Which is not necessarily a problem in itself, as long as you test it 
through and are reasonably sure that it wont break systems en masse. But 
this patch very evidently was not tested in any sufficient manner on its 
primary platform (x86) because my randconfig testsystems (bog standard 
x86 hw) started crashing during bootup almost immediately:

[   22.716747] initcall 0xc0a7b517 ran for 0 msecs: random32_reseed+0x0/0x24()
[   22.722413] Calling initcall 0xc0a7bac1: pci_sysfs_init+0x0/0x44()
[   22.730099] BUG: unable to handle kernel NULL pointer dereference at 0000000c
[   22.737093] IP: [<c034226b>] pcie_aspm_create_sysfs_dev_files+0x1e/0x58
[   22.740003] *pde = 00000000 
[   22.742848] Oops: 0000 [#1] PREEMPT DEBUG_PAGEALLOC
[   22.747699] 
[   22.749174] Pid: 1, comm: swapper Not tainted (2.6.24 #5)
[   22.750000] EIP: 0060:[<c034226b>] EFLAGS: 00010246 CPU: 0
[   22.750000] EIP is at pcie_aspm_create_sysfs_dev_files+0x1e/0x58
[   22.750000] EAX: c32cf504 EBX: c32cf5d0 ECX: 00000000 EDX: 00000000
[   22.750000] ESI: 00000000 EDI: c32cf5d0 EBP: c327df50 ESP: c327df48
[   22.750000]  DS: 007b ES: 007b FS: 0000 GS: 0000 SS: 0068
[   22.750000] Process swapper (pid: 1, ti=c327d000 task=c327c000 task.ti=c327d000)
[   22.750000] Stack: 00000000 00000000 c327df84 c0340c38 c327df74 c0340188 ffffffff c32cf708 
[   22.750000]        00000000 00000000 00000006 c32cf6a8 00000000 c32cf5d0 00000000 c327df94 
[   22.750000]        c0a7badb 00000000 00000000 c327dfe0 c0a6c673 dbc866c8 00000008 00000000 
[   22.750000] Call Trace:
[   22.750000]  [<c0340c38>] ? pci_create_sysfs_dev_files+0x24f/0x2f2
[   22.750000]  [<c0340188>] ? pci_get_subsys+0xc2/0xcb
[   22.750000]  [<c0a7badb>] ? pci_sysfs_init+0x1a/0x44
[   22.750000]  [<c0a6c673>] ? kernel_init+0xb2/0x20b
[   22.750000]  [<c0a6c5c1>] ? kernel_init+0x0/0x20b
[   22.750000]  [<c01057a7>] ? kernel_thread_helper+0x7/0x10
[   22.750000]  =======================
[   22.750000] Code: 84 16 9b c0 e8 2d 6b e4 ff 5b 5e 5d c3 55 89 e5 56 53 89 c3 8b 70 48 f6 80 dd 03 00 00 04 74 41 8a 40 32 3c 04 74 04 3c 06 75 36 <83> 7e 0c 00 74 15 8d 83 38 01 00 00 b9 7c 16 9b c0 ba 68 16 9b 
[   22.750000] EIP: [<c034226b>] pcie_aspm_create_sysfs_dev_files+0x1e/0x58 SS:ESP 0068:c327df48
[   22.750003] ---[ end trace 204172de9b5128e1 ]---

config attached. The crash is fully repeatable on multiple systems. (can 
try any test-patch)

To add insult to injury, the new option is also default-enabled _AND_ 
experimental:

   config PCIEASPM
           bool "PCI Express ASPM support(Experimental)"
           depends on PCI && EXPERIMENTAL
           default y

and it only depends on PCI, so this will crash just about every system 
out there that tracks -git.

please ...

i've punched it out of the randconfig space via the hack below, to keep 
the tests going.

	Ingo

Index: linux/drivers/pci/pcie/Kconfig
===================================================================
--- linux.orig/drivers/pci/pcie/Kconfig
+++ linux/drivers/pci/pcie/Kconfig
@@ -33,6 +33,10 @@ source "drivers/pci/pcie/aer/Kconfig"
 config PCIEASPM
 	bool "PCI Express ASPM support(Experimental)"
 	depends on PCI && EXPERIMENTAL
+
+	# bootup crashes, Sat Feb  2 11:37:58 CET 2008
+	depends on 0
+
 	default y
 	help
 	  This enables PCI Express ASPM (Active State Power Management) and

View attachment "crash.log" of type "text/plain" (267438 bytes)

View attachment "config" of type "text/plain" (44297 bytes)