v2.6.27-rc5 regression: mtd/block device issue with cmd_filter

[Laurent Pinchart]

Hi everybody,

the following commit either introduces a regression or makes and MTD-related issue come up to the surface:

abf5439370491dd6fbb4fe1a7939680d2a9bc9d4 introduces 
block: move cmdfilter from gendisk to request_queue

When registering an MTD device that contains several partitions, the kernel complains about reinitialization of a kobject:

--------------------------- CUT HERE -----------------------
40000000.flash: Found 1 x16 devices at 0x0 in 16-bit bank
Amd/Fujitsu Extended Query Table at 0x0040
40000000.flash: CFI does not contain boot bank location. Assuming top.
number of CFI chips: 1
cfi_cmdset_0002: Disabling erase-suspend-program due to code brokenness.
cmdlinepart partition parsing not available
RedBoot partition parsing not available
Creating 4 MTD partitions on "40000000.flash":                                                                               
0x00000000-0x00040000 : "uboot"
0x00040000-0x00080000 : "firmware"
kobject (c3907194): tried to init an initialized object, something is seriously wrong.
Call Trace:
[c3821c20] [c0008d84] show_stack+0x48/0x16c (unreliable)
[c3821c50] [c01256b0] kobject_init+0x74/0xa0
[c3821c60] [c0125aa4] kobject_init_and_add+0x30/0x78
[c3821d00] [c011d890] blk_register_filter+0x4c/0x8c
[c3821d20] [c011b434] add_disk+0x68/0xc0
[c3821d40] [c0177798] add_mtd_blktrans_dev+0x184/0x294
[c3821d60] [c0178564] mtdblock_add_mtd+0x60/0x80
[c3821d70] [c0177a24] blktrans_notify_add+0x40/0x80
[c3821d90] [c0173730] add_mtd_device+0x9c/0x144
[c3821db0] [c01753d8] add_mtd_partitions+0x4f0/0x62c
[c3821de0] [c0238ec0] 0xc0238ec0
[c3821e40] [c0199060] of_platform_device_probe+0x74/0xa0
[c3821e60] [c01519d4] driver_probe_device+0x120/0x1e4
[c3821e80] [c0151be0] __driver_attach+0x9c/0xd0
[c3821ea0] [c0150618] bus_for_each_dev+0x58/0x9c
[c3821ed0] [c0151c38] driver_attach+0x24/0x34
[c3821ee0] [c0150ff8] bus_add_driver+0x198/0x270
[c3821f10] [c015226c] driver_register+0x64/0x18c
[c3821f30] [c01992d0] of_register_driver+0x44/0x54
[c3821f40] [c02bfcd0] of_flash_init+0x20/0x30
[c3821f50] [c00038f0] do_one_initcall+0x34/0x1c4
[c3821fd0] [c02ac80c] kernel_init+0x94/0x110
[c3821ff0] [c0010328] kernel_thread+0x44/0x60                                                                                
0x00080000-0x00200000 : "kernel"
kobject (c3907194): tried to init an initialized object, something is seriously wrong.
Call Trace:
[c3821c20] [c0008d84] show_stack+0x48/0x16c (unreliable)
[c3821c50] [c01256b0] kobject_init+0x74/0xa0
[c3821c60] [c0125aa4] kobject_init_and_add+0x30/0x78
[c3821d00] [c011d890] blk_register_filter+0x4c/0x8c
[c3821d20] [c011b434] add_disk+0x68/0xc0
[c3821d40] [c0177798] add_mtd_blktrans_dev+0x184/0x294
[c3821d60] [c0178564] mtdblock_add_mtd+0x60/0x80
[c3821d70] [c0177a24] blktrans_notify_add+0x40/0x80
[c3821d90] [c0173730] add_mtd_device+0x9c/0x144
[c3821db0] [c01753d8] add_mtd_partitions+0x4f0/0x62c
[c3821de0] [c0238ec0] 0xc0238ec0
[c3821e40] [c0199060] of_platform_device_probe+0x74/0xa0
[c3821e60] [c01519d4] driver_probe_device+0x120/0x1e4
[c3821e80] [c0151be0] __driver_attach+0x9c/0xd0
[c3821ea0] [c0150618] bus_for_each_dev+0x58/0x9c
[c3821ed0] [c0151c38] driver_attach+0x24/0x34
[c3821ee0] [c0150ff8] bus_add_driver+0x198/0x270
[c3821f10] [c015226c] driver_register+0x64/0x18c
[c3821f30] [c01992d0] of_register_driver+0x44/0x54
[c3821f40] [c02bfcd0] of_flash_init+0x20/0x30
[c3821f50] [c00038f0] do_one_initcall+0x34/0x1c4
[c3821fd0] [c02ac80c] kernel_init+0x94/0x110
[c3821ff0] [c0010328] kernel_thread+0x44/0x60                                                                                
--------------------------- CUT HERE -----------------------

(fourth partition omitted for the sake of clarity, the kobject pointer and stack trace are identical).

The issue is triggered by the MTD block translation layer drivers/mtd/mtdblock.c sharing a single request_queue structure for all its gendisk instances. The queue is allocated and initialized by register_mtd_blktrans in drivers/mtd/mtd_blktrans.c called when the MTD block translation layer is registered, passed to newly created mtd_blktrans_dev in mtdblock_add_mtd through the mtd_blktrans_ops pointer, and set as the gendisk queue in add_mtd_blktrans_dev.

I don't know who is at fault here - the MTD layer for sharing a queue or the block layer for assuming queues are not shared.

Best regards,

-- 
Laurent Pinchart
CSE Semaphore Belgium

Chaussee de Bruxelles, 732A
B-1410 Waterloo
Belgium

T +32 (2) 387 42 59
F +32 (2) 387 42 75
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
Url : http://lists.infradead.org/pipermail/linux-mtd/attachments/20080829/4e17807a/attachment.bin