Warning when mounting btrfs partition, kernel unaligned access

I played a little with btrfs on debian sparc64 (V240). I created a LV from my volumegroup and mounted it. At this moment I got a warning from the kernel as follow:

[  624.857466] device fsid 5defc44f31f449af-b797a325301d119a devid 1 transid 502 /dev/mapper/vgsys-debian64
[  624.983823] ------------[ cut here ]------------
[  625.044657] WARNING: at /build/buildd-linux-2.6_2.6.38-3-sparc-675DHN/linux-2.6-2.6.38/debian/build/source_sparc_none/fs/btrfs/extent_io.c:3787 write_extent_buffer+0xc8/0x128 [btrfs]()
[  625.260939] Modules linked in: nfsd exportfs nfs lockd fscache nfs_acl auth_rpcgss sunrpc ext2 loop flash ext4 mbcache jbd2 crc16 dm_mod raid1 md_mod btrfs lzo_compress zlib_deflate crc32c libcrc32c ide_cd_mod cdrom ata_generic libata ide_pci_generic sd_mod crc_t10dif qla1280 ohci_hcd ehci_hcd tg3 sym53c8xx scsi_transport_spi scsi_mod usbcore libphy alim15x3 nls_base [last unloaded: scsi_wait_scan]
[  625.725351] Call Trace:
[  625.757411]  [00000000103f5cbc] write_extent_buffer+0xc8/0x128 [btrfs]
[  625.843302]  [00000000103fd7b4] btrfs_read_sys_array+0x3c/0x314 [btrfs]
[  625.930333]  [00000000103c65e0] open_ctree+0xf28/0x1884 [btrfs]
[  626.008199]  [000000001039d01c] btrfs_mount+0x238/0x43c [btrfs]
[  626.086068]  [0000000000514db8] vfs_kern_mount+0x98/0x1fc
[  626.157071]  [0000000000514f5c] do_kern_mount+0x24/0xbc
[  626.225696]  [000000000052d718] do_mount+0x830/0x898
[  626.290889]  [000000000054ed68] compat_sys_mount+0x1c8/0x204
[  626.365229]  [0000000000406114] linux_sparc_syscall32+0x34/0x40
[  626.442999] ---[ end trace 1a1c133a476942b9 ]---

Then, after writing on the disk, I got a lot of warning:

[  822.515875] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  825.897439] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  826.030575] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  826.565416] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  826.692390] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  841.631005] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  841.736359] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  846.774635] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  846.880038] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  847.018567] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  847.157098] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  847.262536] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  851.884641] log_unaligned: 127 callbacks suppressed
[  851.948889] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  852.054307] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  852.227997] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]
[  852.333491] Kernel unaligned access at TPC[103c2204] btrfs_csum_final+0x38/0x3c [btrfs]

I peeked a look at the btrf_csum_final and here's the function :
void btrfs_csum_final(u32 crc, char *result)

{
        *(__le32 *)result = ~cpu_to_le32(crc);
}

I know that this problem is certainly an alignment problem and that the sparc is prone to that kind of problem. If someone could explain what's wrong with it?

S. Bernard

On SPARC memory alignment is required. I.e. bytes can be written to any address. 32 bit words on the other hand must be written to an address on a 4 byte boundary, 64 bit words on 8 byte boundaries and so on.

The problem with the csum function is that it casts the result byte pointer to a 32 bits pointer. This pointer might not point to an aligned memory position for 32 bits operations.

Either the function must handle unaligned result pointers internally, or you must guarantee that the result pointer is always aligned. Which would increase performance slightly on most architectures, as those architectures that allow unaligned memory accesses often do it at a cost in memory access time.

Kjetil Oftedal

Hi

=46YI, this has been fixed and is already merged into Linus' tree. Comm= it
7e75bf3ff3a716d7b21d8fb43bf823115801c1e9.

dave
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" = in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

=46rom: David Sterba <[email protected]>
Date: Wed, 13 Apr 2011 11:40:37 +0200

it

Might I suggest adding a BUG_ON() validation of the alignment or similar here?

You can make the test really cheap, and this way no matter what kind of systems the btrfs folks do their testing on this kind of regression will be spotted fast.
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" i= n
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html