01 Aug, 2012

1 commit

  • Add a new operation code (BLKPG_RESIZE_PARTITION) to the BLKPG ioctl that
    allows altering the size of an existing partition, even if it is currently
    in use.

    This patch converts hd_struct->nr_sects into sequence counter because
    One might extend a partition while IO is happening to it and update of
    nr_sects can be non-atomic on 32bit machines with 64bit sector_t. This
    can lead to issues like reading inconsistent size of a partition. Sequence
    counter have been used so that readers don't have to take bdev mutex lock
    as we call sector_in_part() very frequently.

    Now all the access to hd_struct->nr_sects should happen using sequence
    counter read/update helper functions part_nr_sects_read/part_nr_sects_write.
    There is one exception though, set_capacity()/get_capacity(). I think
    theoritically race should exist there too but this patch does not
    modify set_capacity()/get_capacity() due to sheer number of call sites
    and I am afraid that change might break something. I have left that as a
    TODO item. We can handle it later if need be. This patch does not introduce
    any new races as such w.r.t set_capacity()/get_capacity().

    v2: Add CONFIG_LBDAF test to UP preempt case as suggested by Phillip.

    Signed-off-by: Vivek Goyal
    Signed-off-by: Phillip Susi
    Signed-off-by: Jens Axboe

    Vivek Goyal
     

02 Mar, 2012

1 commit

  • Since 2.6.39 (1196f8b), when a driver returns -ENOMEDIUM for open(),
    __blkdev_get() calls rescan_partitions() to remove
    in-kernel partition structures and raise KOBJ_CHANGE uevent.

    However it ends up calling driver's revalidate_disk without open
    and could cause oops.

    In the case of SCSI:

    process A process B
    ----------------------------------------------
    sys_open
    __blkdev_get
    sd_open
    returns -ENOMEDIUM
    scsi_remove_device

    rescan_partitions
    sd_revalidate_disk

    Oopses are reported here:
    http://marc.info/?l=linux-scsi&m=132388619710052

    This patch separates the partition invalidation from rescan_partitions()
    and use it for -ENOMEDIUM case.

    Reported-by: Huajun Li
    Signed-off-by: Jun'ichi Nomura
    Acked-by: Tejun Heo
    Cc: stable@kernel.org
    Signed-off-by: Jens Axboe

    Jun'ichi Nomura
     

04 Jan, 2012

1 commit