Bug#548630: linux-image-2.6.26-2-686: Crash/Hang with heavy disk write when using SATA AHCI mode
- To: Hans IJ <hans.ij@gmail.com>
- Cc: 548630@bugs.debian.org
- Subject: Bug#548630: linux-image-2.6.26-2-686: Crash/Hang with heavy disk write when using SATA AHCI mode
- From: Jonathan Nieder <jrnieder@gmail.com>
- Date: Sat, 3 Sep 2011 12:36:26 -0500
- Message-id: <[🔎] 20110903173549.GA24034@elie>
- Reply-to: Jonathan Nieder <jrnieder@gmail.com>, 548630@bugs.debian.org
- In-reply-to: <b3812d420910151058q6dbc6e44w45bf841ab20b3918@mail.gmail.com>
- References: <b3812d420909290203j26b643afjff47506bd61e7f95@mail.gmail.com> <1255229360.25061.79.camel@localhost> <b3812d420910141304v29da05f2v862b1a620619ccc4@mail.gmail.com> <1255556403.2488.226.camel@localhost> <b3812d420910151058q6dbc6e44w45bf841ab20b3918@mail.gmail.com>
Hi,
Hans IJ wrote:
> I can reproduce this problem on both an ASUS P5Q-VM and Intel DG45ID
> Motherboard. Both MB's have same ICH10 chipset.
> Let me know if you need more.
Sorry for the loong delay. The assertion triggered here is in
mm/vmscan.c, function isolate_lru_pages():
switch (__isolate_lru_page(page, mode, file)) {
case 0:
[...]
case -EBUSY:
[...]
continue;
default:
BUG();
}
eax contains -EINVAL, so chances are that was the error code. It
means the PageLRU flag was clear, the page's mode didn't match "mode",
the page's is_file_cache flag didn't match "file", or the page was
unevictable.
Maybe this is fixed by v2.6.39~59 (mm: check PageUnevictable in
lru_deactivate_fn(), 2011-05-11), even though the case that prompted
that patch was not broken until v2.6.39-rc1~329 (2011-03-22). Can you
still reproduce this? If so, could you test 2.6.39 (which would
work according to that hypothesis) and 2.6.39-rc7 (which wouldn't) to
check?
Hope that helps,
Jonathan
Reply to: