This patch reverts commit f80de881d8df and avoids that sending a
WRITE SAME command to the iSCSI initiator triggers the following:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000014
TARGET_CORE[iSCSI]: Expected Transfer Length: 260096 does not match SCSI CDB
Length: 512 for SAM Opcode: 0x41
IP: iscsi_tcp_segment_done+0x20b/0x310 [libiscsi_tcp]
Oops: 0000 [#1] SMP
Modules linked in: target_core_user uio target_core_iblock target_core_file
iscsi_target_mod target_core_mod netconsole configfs crct10dif_pclmul
crc32_pclmul ghash_clmulni_intel aesni_intel aes_x86_64 crypto_simd cryptd
glue_helper virtio_console virtio_rng virtio_balloon serio_raw i2c_piix4
acpi_cpufreq button iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi ext4
jbd2 mbcache virtio_blk virtio_net psmouse floppy drm_kms_helper syscopyarea
sysfillrect sysimgblt fb_sys_fops ttm drm virtio_pci
CPU: 2 PID: 5 Comm: kworker/u8:0 Not tainted 4.10.0-rc5-debug+ #3
Workqueue: iscsi_q_0 iscsi_xmitworker [libiscsi]
RIP: 0010:iscsi_tcp_segment_done+0x20b/0x310 [libiscsi_tcp]
Call Trace:
iscsi_sw_tcp_xmit_segment+0x84/0x120 [iscsi_tcp]
iscsi_sw_tcp_pdu_xmit+0x51/0x180 [iscsi_tcp]
iscsi_tcp_task_xmit+0xb3/0x290 [libiscsi_tcp]
iscsi_xmit_task+0x4e/0xc0 [libiscsi]
iscsi_xmitworker+0x243/0x330 [libiscsi]
process_one_work+0x1d8/0x4b0
worker_thread+0x49/0x4a0
kthread+0x102/0x140
Fixes: f80de881d8df ("sd: remove __data_len hack for WRITE SAME")
Signed-off-by: Bart Van Assche <bart.vanass...@sandisk.com>
Cc: Christoph Hellwig <h...@lst.de>
Cc: Hannes Reinecke <h...@suse.com>
Cc: Sagi Grimberg <s...@grimberg.me>
Cc: Jens Axboe <ax...@fb.com>
Cc: Lee Duncan <ldun...@suse.com>
Cc: Chris Leech <cle...@redhat.com>
---
drivers/scsi/sd.c | 17 ++++++++++++++++-
1 file changed, 16 insertions(+), 1 deletion(-)
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index 0b09638fa39b..6e6d4a41613f 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -836,6 +836,7 @@ static int sd_setup_write_same_cmnd(struct scsi_cmnd *cmd)
struct bio *bio = rq->bio;
sector_t sector = blk_rq_pos(rq);
unsigned int nr_sectors = blk_rq_sectors(rq);
+ unsigned int nr_bytes = blk_rq_bytes(rq);
int ret;
if (sdkp->device->no_write_same)
@@ -868,7 +869,21 @@ static int sd_setup_write_same_cmnd(struct scsi_cmnd *cmd)
cmd->transfersize = sdp->sector_size;
cmd->allowed = SD_MAX_RETRIES;
- return scsi_init_io(cmd);
+
+ /*
+ * For WRITE SAME the data transferred via the DATA OUT buffer is
+ * different from the amount of data actually written to the target.
+ *
+ * We set up __data_len to the amount of data transferred via the
+ * DATA OUT buffer so that blk_rq_map_sg sets up the proper S/G list
+ * to transfer a single sector of data first, but then reset it to
+ * the amount of data to be written right after so that the I/O path
+ * knows how much to actually write.
+ */
+ rq->__data_len = sdp->sector_size;
+ ret = scsi_init_io(cmd);
+ rq->__data_len = nr_bytes;
+ return ret;
}
static int sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
--
2.11.0
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
