warning during mtd registration in 2.6.27-rc5

[Dmitry Baryshkov]

Hi,

Since commit abf5439370491dd6fbb4fe1a7939680d2a9bc9d4
I'm seeing this problem:

kobject (c3abad68): tried to init an initialized object, something is seriously wrong.
[<c0027ff0>] (dump_stack+0x0/0x14) from [<c013c484>] (kobject_init+0x40/0x78)
[<c013c444>] (kobject_init+0x0/0x78) from [<c013ca24>] (kobject_init_and_add+0x20/0x44)
 r5:c3abad68 r4:c3ac2520
[<c013ca08>] (kobject_init_and_add+0x4/0x44) from [<c0137c6c>] (blk_register_filter+0x48/0x60)
 r5:c3ab4060 r4:c3aba880
[<c0137c24>] (blk_register_filter+0x0/0x60) from [<c01366c4>] (add_disk+0x60/0xb8)
 r4:c3ac2400
[<c0136664>] (add_disk+0x0/0xb8) from [<c01a6994>] (add_mtd_blktrans_dev+0x210/0x244)
 r5:c3ab4060 r4:00000001
[<c01a6784>] (add_mtd_blktrans_dev+0x0/0x244) from [<c01a6bbc>] (mtdblock_add_mtd+0x58/0x64)
 r8:c03c7910 r7:c3ab6100 r6:c03c6bd4 r5:c03c6c04 r4:c3ab6100
[<c01a6b64>] (mtdblock_add_mtd+0x0/0x64) from [<c01a6144>] (blktrans_notify_add+0x3c/0x68)
 r5:c3ab6100 r4:c03c6c04
[<c01a6108>] (blktrans_notify_add+0x0/0x68) from [<c01a3a20>] (add_mtd_device+0xc0/0x120)
 r6:c03c6b80 r5:c3ab6100 r4:c03c6bc4
[<c01a3960>] (add_mtd_device+0x0/0x120) from [<c01a4e50>] (add_one_partition+0x468/0x500)
 r6:c3ac2000 r5:00700000 r4:c3ab6100
[<c01a49e8>] (add_one_partition+0x0/0x500) from [<c01a4f20>] (add_mtd_partitions+0x38/0x7c)
[<c01a4ee8>] (add_mtd_partitions+0x0/0x7c) from [<c0016a48>] (sharpsl_nand_init+0x1dc/0x238)
....

This happens when adding mtd partitions. I've traced it to the fact that
mtd partition devices share the genhd queue with the whole chip block device.
Thus blk_register_filter receives the same pointer for the chip and for each
registered partition.

-- 
With best wishes
Dmitry