Re: [RFC PATCH v2 04/11] pstore: Add compression support to pstore

From: Aruna Balakrishnaiah
Date: Wed Sep 04 2013 - 02:01:16 EST


On Wednesday 04 September 2013 07:14 AM, Seiji Aguchi wrote:
Aruna,

Sorry for the late response.

Seiji,

Could you let us know the efivars buffer size with which the pstore is
registered when
the failure occurred.
I looked into the issue today.

I added some debug message just before pstore_compress().
As you can see below, the buffer size is a fixed value(1024).
Therefore, the size doesn't seem to be related to the failure directly.

Also, in the failure case, zlib_deflate() returns Z_OK and stream.avail_out is zero.
So, I thought big_oops_buf_sz is too big in my environment.
And then, I tuned big_oops_buf_sz to (psinfo->bufsize * 100) / 53.
Seiji,

Thank you for the analysis.

The reason behind compression failure is the size of big_oops_buf which is too
big for efivars case. I will do some experiments with different kind of texts
for buffer size 1024 to check if 100/53 suits for all the cases.


At the same time, while looking into this issue, I had two concerns about current cording.

1) In pstore_decompress(), why not use zlib_inflateInit2() instead of zlib_inflateInit()?
If you use zlib_deflateInit2() for specifying window_bit at compressing time,
zlib_inflateInit2() seems to be appropriate for decompressing.
(Please see a comment about inflateInit2() in include/linux/zlib.h and source code of crypto/deflate.c)

Yes this can be changed to zlib_inflateInit2().

2) As looking at crypto/deflate.c, stream->workspace is kzalloced at the beginning of
compressing/decompressing time.
So, in pstore case, stream.workspace must be initialized to 0 with memset() in pstore_compress()/decompress().

Hmm.. I don't think this is a issue. If you see fs/logfs/compr.c from which I derived the compression
algorithms for pstore as well, they have not initialized stream.workspace. Since the space will be overwritten
during the calculation, I dont think it will matter.

The above 2 things you have suggested are good to have. But the reason behind compression failure is
the big_oops_buf_sz which is too big for efivars.

I did three things above, tuning big_oops_buf_sz, using zlib_inflateInit2() and initializing stream.workspace , in my environment.
As a result, compressions/decmpressions of all entries succeeded with efivars driver.

<snip>
Panic#2 Part1
<4>[ 75.665020] [<ffffffff811af59c>] SyS_write+0x4c/0xa0
<4>[ 75.665020] [<ffffffff8168be82>] system_call_fastpath+0x16/0x1b
<4>[ 75.665020] Code: ef e8 ff f7 ff ff eb d8 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 0f 1f 44 00 00 55 c7 05 cc f3 c9 00 01 00 00 00 48 89 e5 0f ae f8 <c6> 04 25 00 00 00 00 01 5d c3 0f 1f 44 00 00 55 31 c0 c7 05 5e
<1>[ 75.665020] RIP [<ffffffff813d3946>] sysrq_handle_crash+0x16/0x20
<4>[ 75.665020] RSP <ffff88007852de80>
<4>[ 75.665020] CR2: 0000000000000000
<4>[ 75.665020] ---[ end trace 97bb4a1f8d3fe7b2 ]---
<3>[ 75.665020] pstore_dump hsize=13 len=2155 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=13 len=2246 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore: compression failed for Part 2 returned -5
<3>[ 75.665020] pstore: Capture uncompressed oops/panic report of Part 2
<3>[ 75.665020] pstore_dump hsize=13 len=2239 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=13 len=2231 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=13 len=2185 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore: compression failed for Part 5 returned -5
<3>[ 75.665020] pstore: Capture uncompressed oops/panic report of Part 5
<3>[ 75.665020] pstore_dump hsize=13 len=2234 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=13 len=2208 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=13 len=2218 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=13 len=2222 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore: compression failed for Part 9 returned -5
<3>[ 75.665020] pstore: Capture uncompressed oops/panic report of Part 9
<3>[ 75.665020] pstore_dump hsize=14 len=2256 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=14 len=2219 big_oops_buf_sz=2275 psinfo->bufsize=1024
<3>[ 75.665020] pstore_dump hsize=14 len=2226 big_oops_buf_sz=2275 psinfo->bufsize=1024
<0>[ 75.665020] Kernel panic - not syncing: Fatal exception
<3>[ 75.665020] drm_kms_helper: panic occurred, switching back to text console
<snip>

Seiji
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@xxxxxxxxxxxxxxxx
https://lists.ozlabs.org/listinfo/linuxppc-dev


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/