Hi Nicholas,
I also patched rtslib (pull request on github) to work around not
being able to resolve the /dev/* device to the HCL and back the other
way.
[root@tvtln01-eth SOURCES]# head /sys/kernel/config/target/core/pscsi_*/*/info
==> /sys/kernel/config/target/core/pscsi_0/changer1/info <==
Status: ACTIVATED Max Queue Depth: 0 SectorSize: 0 HwMaxSectors: 0
SCSI Device Bus Location: Channel ID: 0 Target ID: 0 LUN: 0 Host ID: 10
Vendor: STK Model: L700 Rev: 0105
==> /sys/kernel/config/target/core/pscsi_1/tape1/info <==
Status: ACTIVATED Max Queue Depth: 0 SectorSize: 0 HwMaxSectors: 0
SCSI Device Bus Location: Channel ID: 0 Target ID: 1 LUN: 0 Host ID: 10
Vendor: IBM Model: ULT3580-TD5 Rev: 0105
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/block_size <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_dpo <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_fua_read <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_fua_write <==
1
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_model_alias
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_rest_reord
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_tas <==
1
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_tpu <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_tpws <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_ua_intlck_ctrl
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/emulate_write_cache
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/enforce_pr_isids <==
1
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/fabric_max_sectors
<==
8192
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/hw_block_size <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/hw_max_sectors <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/hw_queue_depth <==
32
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/is_nonrot <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/max_unmap_block_desc_count
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/max_unmap_lba_count
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/max_write_same_len
<==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/optimal_sectors <==
8192
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/queue_depth <==
32
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/unmap_granularity <==
0
==> /sys/kernel/config/target/core/pscsi_0/changer1/attrib/unmap_granularity_alignment
<==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/block_size <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_dpo <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_fua_read <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_fua_write <==
1
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_model_alias <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_rest_reord <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_tas <==
1
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_tpu <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_tpws <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_ua_intlck_ctrl
<==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/emulate_write_cache <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/enforce_pr_isids <==
1
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/fabric_max_sectors <==
8192
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/hw_block_size <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/hw_max_sectors <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/hw_queue_depth <==
32
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/is_nonrot <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/max_unmap_block_desc_count
<==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/max_unmap_lba_count <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/max_write_same_len <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/optimal_sectors <==
8192
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/queue_depth <==
32
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/unmap_granularity <==
0
==> /sys/kernel/config/target/core/pscsi_1/tape1/attrib/unmap_granularity_alignment
<==
0
Here is my patch below. I went with <1 because I was too lazy to check
if the value was signed or not, so it can probably be changed to an =
It's not a big patch at all but it works quite well!
*** a/drivers/target/target_core_device.c 2015-01-29
18:15:53.000000000 -0500
--- b/drivers/target/target_core_device.c.mod 2016-10-30
03:21:41.955410797 -0400
***************
*** 639,644 ****
--- 639,647 ----
* Limit max_sectors to a PAGE_SIZE aligned value for modern
* transport_allocate_data_tasks() operation.
*/
+
+ if (block_size < 1)
+ return 0;
alignment = max(1ul, PAGE_SIZE / block_size);
aligned_max_sectors = rounddown(max_sectors, alignment);
On Mon, Oct 31, 2016 at 5:00 PM, Nicholas A. Bellinger
<nab@xxxxxxxxxxxxxxx> wrote:
> Hi Malcolm,
>
> On Mon, 2016-10-31 at 12:16 +1100, Malcolm Haak wrote:
>> Hi all,
>>
>> In my quest to get targetcli exporting mhvtl devices I ran into a dev0
>> in se_dev_align_max_sectors.
>>
>
> Thanks for reporting this long standing PSCSI w/ TYPE_TAPE bug.
>
>> Now I am running kernel-3.10.0-229.el7 so it might be patched in newer
>> kernels, but it seems that when using pscsi to export tape/changer
>> scsi devices (being that they aren't block devices) it encounters a
>> zero value for block_size.
>>
>> I've made my self a patch that sets a zero block size to PAGE_SIZE and
>> it's working, but I'm not sure if this is the correct fix. (Patch is
>> just a simple if block_size == 0 then block_size = PAGE_SIZE)
>>
>
> Sure, let's have a look. Feel free to post your patch here.
>
> Btw, how do your PSCSI backend settings look..?
>
> head /sys/kernel/config/target/core/pscsi_*/*/info
> head /sys/kernel/config/target/core/pscsi_*/*/attrib/*
>
>> So I guess I'm trying to a) find out if that is a valid fix and b)
>> regardless if it is/isn't get a fix into mainline for other people to
>> use.
>>
>>
>
> IIRC, there is an ancient hack in pscsi_transport_complete() to set the
> correct sd->sector_size during MODE_SELECT from the underlying
> scsi_device. Once upon a time ~10 years ago this was required to get
> things working..
>
> I'd be amazed if we still need that hack, but as you've noticed it's
> been awhile since someone made PSCSI w/ TYPE_TAPE export work.
>
> Nice work btw.
>
--
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
