[OpenAFS-devel] Bug with bypass-cache

Hartmut Reuter reuter@rzg.mpg.de
Mon, 10 Jan 2011 15:05:43 +0100


Ever someone tried bypass caching? I have my doubts!

It's probably not the best idea to use for
FCSDesireBypass the same bit as for CStatd and for
FCSBypass the same bit as for CBackup and for
FCSManuallySet the same as for CRO.

If you try to read a file bypassing the cache you get

[ 2103.337408] openafs: afs_TransitionToBypass: illegal transition to 
bypass--already FCSBypass
[ 2103.337558] ------------[ cut here ]------------
[ 2103.337598] kernel BUG at 
/usr/src/packages/SOURCES/openafs-1.6.0pre1/src/libafs/MODLOAD-2.6.31.14-0.6-default-MP/afs_bypasscache.c:147!
[ 2103.337669] invalid opcode: 0000 [#1] SMP
[ 2103.337709] last sysfs file: 
/sys/devices/LNXSYSTM:00/device:00/PNP0A08:00/device:02/PNP0C0A:00/power_supply/BAT0/charge_full
[ 2103.337774] Modules linked in: libafs(P) i915 drm i2c_algo_bit snd_pcm_oss 
snd_mixer_oss snd_seq snd_seq_device edd ipv6 af_packet cpufreq_conservative 
cpufreq_userspace cpufreq_powersave acpi_cpufreq speedstep_lib binfmt_misc fuse 
xfs exportfs loop dm_mod arc4 ecb snd_hda_codec_realtek cryptomgr aead pcompress 
crypto_blkcipher crypto_hash snd_hda_intel crypto_algapi snd_hda_codec ath9k 
mac80211 snd_hwdep uvcvideo ath snd_pcm snd_timer videodev snd v4l1_compat 
cfg80211 video soundcore iTCO_wdt i2c_i801 iTCO_vendor_support led_class sg 
snd_page_alloc i2c_core rfkill pcspkr atl1c output joydev intel_agp battery ac 
button wmi ext4 jbd2 crc16 ahci rtc_cmos uhci_hcd libata rtc_core rtc_lib 
ehci_hcd fan thermal processor thermal_sys hwmon [lastunloaded: libafs]
[ 2103.338531]
[ 2103.338560] Pid: 8015, comm: cat Tainted: P           (2.6.31.14-0.6-default 
#1) Aspire one
[ 2103.338612] EIP: 0060:[<f8370661>] EFLAGS: 00010246 CPU: 1
[ 2103.338726] EIP is at afs_TransitionToBypass+0x251/0x260 [libafs]
[ 2103.338768] EAX: 00000066 EBX: d6608840 ECX: ffffff9d EDX: 00000000
[ 2103.338808] ESI: 00000001 EDI: d633f840 EBP: e8221bb4 ESP: e8221b74
[ 2103.338849]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[ 2103.338889] Process cat (pid: 8015, ti=e8220000 task=f55de0b0 task.ti=e8220000)
[ 2103.338931] Stack:
[ 2103.338952]  f83d5898 f83700ca 00000000 d633f840 e8221ba0 c0232e6f fffff000 
e8221bac
[ 2103.339044] <0> e8221bac c0232d27 825e2dc4 fffa0000 825e2dc4 d633f840 
00000001 00000004
[ 2103.339126] <0> e8221ca0 f83c5daa 00000001 c17acd40 0000002c 00000000 
c08a2e64 00000002
[ 2103.339217] Call Trace:
[ 2103.339589]  [<f83c5daa>] afs_linux_readpages+0xb9a/0xe00 [libafs]
[ 2103.340164]  [<c02e145e>] read_pages+0x3e/0xf0
[ 2103.340222]  [<c02e168c>] __do_page_cache_readahead+0x17c/0x1b0
[ 2103.340276]  [<c02e16ed>] ra_submit+0x2d/0x50
[ 2103.340324]  [<c02e18c3>] ondemand_readahead+0x93/0x1f0
[ 2103.340375]  [<c02e1abb>] page_cache_async_readahead+0x9b/0xb0
[ 2103.340429]  [<c02d94be>] T.875+0x21e/0x470
[ 2103.340477]  [<c02d97c2>] generic_file_aio_read+0xb2/0x250
[ 2103.340546]  [<c031385c>] do_sync_read+0xdc/0x130
[ 2103.340714]  [<f83c3509>] afs_linux_read+0xb9/0x270 [libafs]
[ 2103.341124]  [<c0313da7>] vfs_read+0xb7/0x1b0
[ 2103.341404]  [<c0314513>] sys_read+0x53/0xa0
[ 2103.341404]  [<c0203f43>] sysenter_do_call+0x12/0x28
[ 2103.341404]  [<ffffe430>] 0xffffe430
[ 2103.341404] Code: 44 24 08 c9 00 00 00 c7 44 24 04 78 57 3d f8 c7 04 24 e4 57 
3d f8 e8 fd 02 29 c8 0f 0b eb fe c7 04 24 98 58 3d f8 e8 ed 02 29 c8 <0f> 0b eb 
fe 8d 74 26 00 8d bc 27 0000 00 00 55 89 e5 57 56 53
[ 2103.341806] EIP: [<f8370661>] afs_TransitionToBypass+0x251/0x260 [libafs] 
SS:ESP 0068:e8221b74
[ 2103.342234] ---[ end trace 471facebb05395a1 ]---

cmdebug shows consequently for this file:

** Cache entry @ 0xd6608840 for 195.537009025.3698.559291 [notebook]
        100000000 bytes  DV            8  refcnt     1
     callback e80be440   expires 1294668805
     0 opens     0 writers
     normal file
     states (0x3), stat'd, backup

even if it is the RW-volume.

Hartmut

Derrick Brashear wrote:
> On Fri, Jan 7, 2011 at 4:22 PM, Rich Sudlow<rich@nd.edu>  wrote:
>> Derrick J Brashear wrote:
>>>
>>> The OpenAFS Gatekeepers announce the availability of the first release
>>> candidate for OpenAFS version 1.6.0.
>>> Source files and available binaries can be accessed via the web at:
>>>
>>>        http://www.openafs.org/release/openafs-1.6.0pre1.html
>>>
>>> or via AFS at:
>>>
>>>        UNIX: /afs/grand.central.org/software/openafs/candidate/1.6.0pre1/
>>>        UNC:  \\afs\grand.central.org\software\openafs\candidate\1.6.0pre1\
>>>
>>> A large number of bugfixes, a number of new features, and a set of
>>> binaries supporting the Demand Attach File Service, are included.
>>>
>>> Please assist the gatekeepers by deploying this release and providing
>>> positive or negative feedback. Bug reports should be filed to
>>> openafs-bugs@openafs.org .  Reports of success should be sent to
>>> openafs-info@openafs.org .
>>>
>>> Derrick Brashear
>>> for the OpenAFS Gatekeepers
>>> _______________________________________________
>>> OpenAFS-announce mailing list
>>> OpenAFS-announce@openafs.org
>>> https://lists.openafs.org/mailman/listinfo/openafs-announce
>>
>> Should I expect the source rpm to work with Red Hat 6?
>
> Given what I know at the moment some tweaking is going to be needed
> before it does.
>
>


-- 
-----------------------------------------------------------------
Hartmut Reuter                  e-mail 		reuter@rzg.mpg.de
			   	phone 		 +49-89-3299-1328
			   	fax   		 +49-89-3299-1301
RZG (Rechenzentrum Garching)   	web    http://www.rzg.mpg.de/~hwr
Computing Center of the Max-Planck-Gesellschaft (MPG) and the
Institut fuer Plasmaphysik (IPP)
-----------------------------------------------------------------