Il 04/09/2012 22:25, Nicholas A. Bellinger ha scritto:
> On Tue, 2012-09-04 at 21:44 +0200, Paolo Bonzini wrote:
>> So far I reproduced this with 3.4.0. I'll try a more recent kernel tomorrow.
>>
>> It reproduces with both loopback and iSCSI backends. Here is it with loopback.
>
> Looks like we are overflowing somewhere during the end of lba check for
> this case. Btw, this code is generic to all backends for data I/O, so
> I'm thinking it's likely a bug that is independent of pSCSI.
More information... this time with 3.6.0-rc4:
* I cannot reproduce it anymore with loopback.
* with iSCSI I get the right asc/ascq but the first 8 bytes of the sense
are all wrong.
Example:
expected: 70 00 05 00 00 00 00 0a 00 00 00 00 21 00 00 00 00 00
wrong: 00 f3 5e 2d 00 88 ff ff 00 00 00 00 21 00 00 00 00 00
wrong: c0 73 7a 2d 00 88 ff ff 00 00 00 00 21 00 00 00 00 00
^^^^^^^^ ^^^^^^^^^^^^^^
this tends to this seems to be
remain the same always the same
across an iSCSI
session
These 8 bytes seem to depend on the target configuration (!). The above
is with 2xPSCSI (sda and sr0). With 1xIBLOCK + 1xPSCSI, I get all
zeroes in the 8 bytes. Probably just a quirk, though.
* with 3.6.0-rc4, it does not reproduce with IBLOCK. However, IBLOCK
gives a wrong asc (0x24/0x00, i.e. invalid field in CDB).
BTW, the reproducer is the iSCSI test tool at
github.com/bonzini/libiscsi (that's just my fork, the author is Ronnie
Sahlberg). Run "bin/iscsi-test -tT0101_read10_beyond_eol
iscsi://localhost:3260/$IQN/$LUN".
Paolo
--
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
