Re: current state of the great stack overflow hunt

On Sun, Jan 18, 2009 at 11:48 AM, John Hughes <[email protected]> wrote:
> John Hughes wrote:
>> do_IRQ: stack overflow: 328
>>  [<c01079ee>] dump_stack+0x1e/0x20
>>  [<c01094ae>] __overflow+0x1e/0x20
>>  [<c010946c>] do_IRQ+0xac/0xd0
>>  =======================
>>  ...
>>  [<d08ea8dd>] drbd_make_request_26+0xed/0x2c0 [drbd]
>>  [<c038847a>] generic_make_request+0xca/0x270
>>  ...
>>  [<d0958142>] ext3_prepare_write+0xf2/0x1a0 [ext3]
>>  [<c0152452>] generic_file_buffered_write+0x202/0x720
>>  [<c0152c4b>] __generic_file_aio_write_nolock+0x2db/0x510
>>  [<c0152fc0>] __generic_file_write_nolock+0x90/0xb0
>>  [<c0153370>] generic_file_writev+0x50/0xe0
>>  [<c02d9c20>] cfsd_write+0x230/0x290
>>  ...
>>  [<c02d0a69>] cfs_sync_inode+0x59/0x90
>>  [<c02cd1e6>] cfs_file_flush+0x96/0x100
>>  [<c017023d>] filp_close+0x6d/0x90
>>  [<c01702cb>] sys_close+0x6b/0x90
>>  [<c0106997>] sysenter_past_esp+0x54/0x75
>> Back from overflow
>>
> So the problem is the traditional device stacking problem, and we only
> have around 364 bytes (<10% of the stack) in cfs functions.
>
> Seems to me that there's not much we can do about this, we'll need
> whatever stack reduction that the base kernel and drbd have managed to
> get in newer versions (of the kernel and drbd).
>
> Roger - what, if any, changes have to be made to drbd 8.2 to make it
> work with OpenSSI?
>

I'd imagine allocating more structures from memory to save <<10% to
support 4K stacks actually increases fragmentation and goes against
the purpose of 4K stacks.

Also according to this article (http://kerneltrap.org/8k_stacks) the
base kernel with 4K stacks already runs into problems with some
configurations - eg. nfs+xfs+raid.  It's not worth chasing drbd.

There is something we can do and that is 64-bit OpenSSI.

------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword