cxlflash: Fix to drain operations from previous reset
authorManoj N. Kumar <manoj@linux.vnet.ibm.com>
Wed, 15 Jun 2016 23:49:20 +0000 (18:49 -0500)
committerMartin K. Petersen <martin.petersen@oracle.com>
Wed, 13 Jul 2016 03:16:31 +0000 (23:16 -0400)
commitf411396d9492761ab4f9fb58aa363a2926b94af8
tree244a3820fd8b2d88eb8ce6d736c53fb74995d50b
parent2a81ffdd9da1dc122c195cd53a0c1605f29c10d0
cxlflash: Fix to drain operations from previous reset

While running 'sg_reset -H' in a loop with a user-space application active,
hit the following exception:

cpu 0x2: Vector: 300 (Data Access)
    pc: : afu_attach+0x50/0x240 [cxlflash]
    lr: : cxlflash_afu_recover+0x3dc/0x7d0 [cxlflash]
    pid   = 20365, comm = run_block_fvt

Linux version 4.5.0-491-26f710d+

cxlflash_afu_recover+0x3dc/0x7d0 [cxlflash]
cxlflash_ioctl+0x5a8/0x6f0 [cxlflash]
scsi_ioctl+0x3b0/0x4c0
sd_ioctl+0x110/0x190
blkdev_ioctl+0x28c/0xc20
block_ioctl+0xa4/0xd0
do_vfs_ioctl+0xd8/0x8c0
SyS_ioctl+0xd4/0xf0
system_call+0x38/0xb4

The problem here is that the problem space area is unmapped while the
application issues the DK_CXLFLASH_RECOVER_AFU ioctl.

This is the order I observe:

proc1 proc2
1) sg_reset
2) ioctl(DK_CXLFLASH_RECOVER_AFU)
3) sg_reset again
   causing a PSA unmap
4) continues RECOVER_AFU processing

The resolution to this problem is to have the reset handler drain all
outstanding user space initiated ioctls before proceeding.  It is safe
to drain after the state has been changed to STATE_RESET. Also since
drain_ioctls() was static, it had to be moved up a bit to be before
cxlflash_eh_host_reset_handler().

Signed-off-by: Manoj N. Kumar <manoj@linux.vnet.ibm.com>
Acked-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/cxlflash/main.c