Hi Sebastian,
On Fri, 2012-11-30 at 10:50 +0100, Sebastian Andrzej Siewior wrote:
> nab, your next tree is broken. I have here file backend setup with
> tcm_loopback. Since commit 8f9f44f8 in your for-next branch of
> git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
>
So far I've been trying to reproduce with for-next + master code with a
FILEIO + xfs on top, but am not having any luck triggering this bug as
of yet.
> I get
>
> SCSI subsystem initialized
> Rounding down aligned max_sectors from 4294967295 to 8388600
> scsi0 : TCM_Loopback
> scsi 0:0:1:0: Direct-Access LIO-ORG FILEIO 4.0 PQ: 0
> ANSI: 5
> sd 0:0:1:0: [sda] 61441 512-byte logical blocks: (31.4 MB/30.0 MiB)
> sd 0:0:1:0: [sda] Write Protect is off
> sd 0:0:1:0: [sda] Mode Sense: 2f 00 00 00
> sd 0:0:1:0: [sda] Write cache: disabled, read cache: enabled, doesn't
> support DPO or FUA
> sda: sda1
> sd 0:0:1:0: [sda] Attached SCSI disk
> sd 0:0:1:0: Attached scsi generic sg0 type 0
> SGI XFS with security attributes, no debug enabled
> XFS (sda1): Mounting Filesystem
> XFS (sda1): Ending clean mount
> vfs_readv() returned -14 for non S_ISBLK
> sd 0:0:1:0: [sda]
> Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
> sd 0:0:1:0: [sda]
> Sense Key : Illegal Request [current]
> sd 0:0:1:0: [sda]
> <<vendor>> ASC=0x80 ASCQ=0x0ASC=0x80 ASCQ=0x0
> sd 0:0:1:0: [sda] CDB:
> Read(10): 28 00 00 00 30 3d 00 00 80 00
> end_request: I/O error, dev sda, sector 12349
Ok, this looks like a perfectly normal READ_10 of 64K (128 sectors) to
LBA 12349.
Not sure why vfs_readv() of 64K is generating an -EFAULT here, but lets
first verify that SGLs being dispatched by tcm_loop to TCM -> FILEIO are
looking as expected on your setup.
Please try to reproduce with the following patch:
diff --git a/drivers/target/target_core_file.c b/drivers/target/target_core_file.c
index 8f6dae0..0a46137 100644
--- a/drivers/target/target_core_file.c
+++ b/drivers/target/target_core_file.c
@@ -261,6 +261,11 @@ static int fd_do_readv(struct se_cmd *cmd, struct scatterlist *sgl,
if (ret < 0) {
pr_err("vfs_readv() returned %d for non"
" S_ISBLK\n", ret);
+
+ for_each_sg(sgl, sg, sgl_nents, i) {
+ pr_err("SG[%d]: %p page: %p length: %u offset: %d\n",
+ i, sg, sg_page(sg), sg->length, sg->offset);
+ }
return ret;
}
}
> vfs_readv() returned -14 for non S_ISBLK
> sd 0:0:1:0: [sda]
> Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
> sd 0:0:1:0: [sda]
> Sense Key : Illegal Request [current]
> sd 0:0:1:0: [sda]
> <<vendor>> ASC=0x80 ASCQ=0x0ASC=0x80 ASCQ=0x0
> sd 0:0:1:0: [sda] CDB:
> Read(10): 28 00 00 00 30 3d 00 00 08 00
> end_request: I/O error, dev sda, sector 12349
> sha1sum: /mnt/sdb/libc-2.11.2.so: Input/output error
>
> It wirks one commit before the commit I mentioned. The commit in
> question is:
>
> commit 8f9f44f8957b262de717a48269a5ceca36c2b504
> Author: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx>
> Date: Mon Oct 1 23:29:49 2012 -0700
>
> tcm_loop: Convert I/O path to use target_submit_cmd_map_sgls
>
> This patch converts tcm_loop to use target_submit_cmd_map_sgls() for
> I/O submission and mapping of pre-allocated SGL memory from incoming
> scsi_cmnd -> se_cmd descriptors.
>
> This includes removing the original open-coded fabric uses of target
> core callers to support transport_generic_map_mem_to_cmd() between
> target_setup_cmd_from_cdb() and transport_handle_cdb_direct() logic.
>
> (v2: Use renamed target_submit_cmd_map_sgls)
>
> Reported-by: Christoph Hellwig <hch@xxxxxx>
> Reviewed-by: Christoph Hellwig <hch@xxxxxx>
> Signed-off-by: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx>
>
Very strange. This is a non functional change for tcm_loop code using
the exact same underlying target_core_mod functions for se_cmd setup ->
SGL mapping -> se_cmd I/O dispatch ub target_submit_cmd_map_sg(), so I
don't see how it would effect this bug.
Thanks for reporting!
--nab
--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
