* [PATCH] scsi_debug: Fix endianess in partition table
@ 2013-02-11 17:34 Martin Peschke
2013-02-12 9:03 ` Martin Peschke
2013-02-13 20:08 ` Douglas Gilbert
0 siblings, 2 replies; 6+ messages in thread
From: Martin Peschke @ 2013-02-11 17:34 UTC (permalink / raw)
To: linux-scsi
Both start_sect and nr_sects in struct partition are __le32 and
require cpu_to_le32() on assignment.
Without this fix tools like fdisk show an invalid partition table
for SCSI devices emulated by scsi_debug on big-endian architectures,
like s390x. Besides a kernel message like this was emitted:
sda: p1 start 536870912 is beyond EOD, enabling native capacity
sda: p1 start 536870912 is beyond EOD, truncated
For verification 'xxd -l 512 /dev/sda' has been used to make sure
that this fix makes scsi_debug generated partition tables on s390x
look like the ones generated on my laptop.
Signed-off-by: Martin Peschke <mpeschke@linux.vnet.ibm.com>
Reviewed-by: Steffen Maier <maier@linux.vnet.ibm.com>
---
drivers/scsi/scsi_debug.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -2662,8 +2662,8 @@ static void __init sdebug_build_parts(un
/ sdebug_sectors_per;
pp->end_sector = (end_sec % sdebug_sectors_per) + 1;
- pp->start_sect = start_sec;
- pp->nr_sects = end_sec - start_sec + 1;
+ pp->start_sect = cpu_to_le32(start_sec);
+ pp->nr_sects = cpu_to_le32(end_sec - start_sec + 1);
pp->sys_ind = 0x83; /* plain Linux partition */
}
}
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] scsi_debug: Fix endianess in partition table
2013-02-11 17:34 [PATCH] scsi_debug: Fix endianess in partition table Martin Peschke
@ 2013-02-12 9:03 ` Martin Peschke
2013-02-12 14:45 ` Douglas Gilbert
2013-02-13 20:08 ` Douglas Gilbert
1 sibling, 1 reply; 6+ messages in thread
From: Martin Peschke @ 2013-02-12 9:03 UTC (permalink / raw)
To: linux-scsi
On Mon, 2013-02-11 at 18:34 +0100, Martin Peschke wrote:
> Both start_sect and nr_sects in struct partition are __le32 and
> require cpu_to_le32() on assignment.
Steffen Maier has pointed me at:
block/partitions/msdos.c: return
(sector_t)get_unaligned_le32(&p->start_sect);
Unfortunately, both get_unaligned_le32() and le32_to_cpu() appear to be
in use for start_sect and nr_sects.
Any one who would argue for changing my patch from cpu_to_le32 to
put_unaligned_le32()?
Thanks,
Martin
>
> Without this fix tools like fdisk show an invalid partition table
> for SCSI devices emulated by scsi_debug on big-endian architectures,
> like s390x. Besides a kernel message like this was emitted:
>
> sda: p1 start 536870912 is beyond EOD, enabling native capacity
> sda: p1 start 536870912 is beyond EOD, truncated
>
> For verification 'xxd -l 512 /dev/sda' has been used to make sure
> that this fix makes scsi_debug generated partition tables on s390x
> look like the ones generated on my laptop.
>
> Signed-off-by: Martin Peschke <mpeschke@linux.vnet.ibm.com>
> Reviewed-by: Steffen Maier <maier@linux.vnet.ibm.com>
>
> ---
> drivers/scsi/scsi_debug.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> --- a/drivers/scsi/scsi_debug.c
> +++ b/drivers/scsi/scsi_debug.c
> @@ -2662,8 +2662,8 @@ static void __init sdebug_build_parts(un
> / sdebug_sectors_per;
> pp->end_sector = (end_sec % sdebug_sectors_per) + 1;
>
> - pp->start_sect = start_sec;
> - pp->nr_sects = end_sec - start_sec + 1;
> + pp->start_sect = cpu_to_le32(start_sec);
> + pp->nr_sects = cpu_to_le32(end_sec - start_sec + 1);
> pp->sys_ind = 0x83; /* plain Linux partition */
> }
> }
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] scsi_debug: Fix endianess in partition table
2013-02-12 9:03 ` Martin Peschke
@ 2013-02-12 14:45 ` Douglas Gilbert
2013-02-13 15:16 ` Martin Peschke
0 siblings, 1 reply; 6+ messages in thread
From: Douglas Gilbert @ 2013-02-12 14:45 UTC (permalink / raw)
To: Martin Peschke; +Cc: linux-scsi
On 13-02-12 04:03 AM, Martin Peschke wrote:
> On Mon, 2013-02-11 at 18:34 +0100, Martin Peschke wrote:
>> Both start_sect and nr_sects in struct partition are __le32 and
>> require cpu_to_le32() on assignment.
>
> Steffen Maier has pointed me at:
>
> block/partitions/msdos.c: return
> (sector_t)get_unaligned_le32(&p->start_sect);
>
> Unfortunately, both get_unaligned_le32() and le32_to_cpu() appear to be
> in use for start_sect and nr_sects.
>
> Any one who would argue for changing my patch from cpu_to_le32 to
> put_unaligned_le32()?
No (because I don't know). However since SCSI is big
endian and you are introducing some "le" code then a line
or so of explanation (comments) in your revised patch might
be helpful.
BTW Finding a big endian architecture to test this patch on is
not easy. The openwrt in my router is big endian (MIPS) but
openwrt don't distribute the scsi_debug module :-(
Doug Gilbert
>> Without this fix tools like fdisk show an invalid partition table
>> for SCSI devices emulated by scsi_debug on big-endian architectures,
>> like s390x. Besides a kernel message like this was emitted:
>>
>> sda: p1 start 536870912 is beyond EOD, enabling native capacity
>> sda: p1 start 536870912 is beyond EOD, truncated
>>
>> For verification 'xxd -l 512 /dev/sda' has been used to make sure
>> that this fix makes scsi_debug generated partition tables on s390x
>> look like the ones generated on my laptop.
>>
>> Signed-off-by: Martin Peschke <mpeschke@linux.vnet.ibm.com>
>> Reviewed-by: Steffen Maier <maier@linux.vnet.ibm.com>
>>
>> ---
>> drivers/scsi/scsi_debug.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> --- a/drivers/scsi/scsi_debug.c
>> +++ b/drivers/scsi/scsi_debug.c
>> @@ -2662,8 +2662,8 @@ static void __init sdebug_build_parts(un
>> / sdebug_sectors_per;
>> pp->end_sector = (end_sec % sdebug_sectors_per) + 1;
>>
>> - pp->start_sect = start_sec;
>> - pp->nr_sects = end_sec - start_sec + 1;
>> + pp->start_sect = cpu_to_le32(start_sec);
>> + pp->nr_sects = cpu_to_le32(end_sec - start_sec + 1);
>> pp->sys_ind = 0x83; /* plain Linux partition */
>> }
>> }
>>
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] scsi_debug: Fix endianess in partition table
2013-02-12 14:45 ` Douglas Gilbert
@ 2013-02-13 15:16 ` Martin Peschke
0 siblings, 0 replies; 6+ messages in thread
From: Martin Peschke @ 2013-02-13 15:16 UTC (permalink / raw)
To: dgilbert; +Cc: linux-scsi
On Tue, 2013-02-12 at 09:45 -0500, Douglas Gilbert wrote:
> However since SCSI is big endian and you are introducing
> some "le" code then a line or so of explanation (comments)
> in your revised patch might be helpful.
I would argue that the definition of struct partition itself should be
sufficient explanation. Only scsi_debug failed to assign values in a
correct manner.
> BTW Finding a big endian architecture to test this patch on is
> not easy.
No big deal ;-) I can help out with test data from my System z (aka
s390x).
Without fix:
[root@zzzzzzzz ~]# xxd -l 512 /dev/sda
<snip>
00001b0: 0000 0000 0000 0000 0000 0000 0000 0001 ................
00001c0: 0100 8307 203f 0000 0020 0000 3fe0 0000 .... ?... ..?...
00001d0: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00001e0: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00001f0: 0000 0000 0000 0000 0000 0000 0000 55aa ..............U.
With fix:
[root@zzzzzzzz ~]# xxd -l 512 /dev/sda
<snip>
00001b0: 0000 0000 0000 0000 0000 0000 0000 0001 ................
00001c0: 0100 8307 203f 2000 0000 e03f 0000 0000 .... ? ....?....
00001d0: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00001e0: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00001f0: 0000 0000 0000 0000 0000 0000 0000 55aa ..............U.
Which makes it exactly look like the table seen on my laptop:
root@xxxxxxxx:~# xxd -l 512 /dev/sdb
<snip>
00001b0: 0000 0000 0000 0000 0000 0000 0000 0001 ................
00001c0: 0100 8307 203f 2000 0000 e03f 0000 0000 .... ? ....?....
00001d0: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00001e0: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00001f0: 0000 0000 0000 0000 0000 0000 0000 55aa ..............U.
Assuming:
modprobe scsi_debug physblk_exp=3 lowest_aligned=7 num_parts=1
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] scsi_debug: Fix endianess in partition table
2013-02-11 17:34 [PATCH] scsi_debug: Fix endianess in partition table Martin Peschke
2013-02-12 9:03 ` Martin Peschke
@ 2013-02-13 20:08 ` Douglas Gilbert
1 sibling, 0 replies; 6+ messages in thread
From: Douglas Gilbert @ 2013-02-13 20:08 UTC (permalink / raw)
To: Martin Peschke; +Cc: linux-scsi
On 13-02-11 12:34 PM, Martin Peschke wrote:
> Both start_sect and nr_sects in struct partition are __le32 and
> require cpu_to_le32() on assignment.
>
> Without this fix tools like fdisk show an invalid partition table
> for SCSI devices emulated by scsi_debug on big-endian architectures,
> like s390x. Besides a kernel message like this was emitted:
>
> sda: p1 start 536870912 is beyond EOD, enabling native capacity
> sda: p1 start 536870912 is beyond EOD, truncated
>
> For verification 'xxd -l 512 /dev/sda' has been used to make sure
> that this fix makes scsi_debug generated partition tables on s390x
> look like the ones generated on my laptop.
>
> Signed-off-by: Martin Peschke <mpeschke@linux.vnet.ibm.com>
> Reviewed-by: Steffen Maier <maier@linux.vnet.ibm.com>
Acked-by: Douglas Gilbert <dgilbert@interlog.com>
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH] scsi_debug: Fix endianess in partition table
@ 2013-02-15 12:30 Martin Peschke
0 siblings, 0 replies; 6+ messages in thread
From: Martin Peschke @ 2013-02-15 12:30 UTC (permalink / raw)
To: James.Bottomley; +Cc: linux-scsi
James,
could you pick up this fix, please?
Btw., it looks like others where bitten by this one:
https://nazar.karan.org/blob/distro!
parted.git/8780c767126938173f49dfbcd4360813932f7756/SOURCES!
disable-t9020.patch
Thanks,
Martin
Both start_sect and nr_sects in struct partition are __le32 and
require cpu_to_le32() on assignment.
Without this fix tools like fdisk show an invalid partition table
for SCSI devices emulated by scsi_debug on big-endian architectures,
like s390x. Besides a kernel message like this was emitted:
sda: p1 start 536870912 is beyond EOD, enabling native capacity
sda: p1 start 536870912 is beyond EOD, truncated
For verification 'xxd -l 512 /dev/sda' has been used to make sure
that this fix makes scsi_debug generated partition tables on s390x
look like the ones generated on my laptop.
Signed-off-by: Martin Peschke <mpeschke@linux.vnet.ibm.com>
Reviewed-by: Steffen Maier <maier@linux.vnet.ibm.com>
Acked-by: Douglas Gilbert <dgilbert@interlog.com>
---
drivers/scsi/scsi_debug.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -2662,8 +2662,8 @@ static void __init sdebug_build_parts(un
/ sdebug_sectors_per;
pp->end_sector = (end_sec % sdebug_sectors_per) + 1;
- pp->start_sect = start_sec;
- pp->nr_sects = end_sec - start_sec + 1;
+ pp->start_sect = cpu_to_le32(start_sec);
+ pp->nr_sects = cpu_to_le32(end_sec - start_sec + 1);
pp->sys_ind = 0x83; /* plain Linux partition */
}
}
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2013-02-15 12:30 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-02-11 17:34 [PATCH] scsi_debug: Fix endianess in partition table Martin Peschke
2013-02-12 9:03 ` Martin Peschke
2013-02-12 14:45 ` Douglas Gilbert
2013-02-13 15:16 ` Martin Peschke
2013-02-13 20:08 ` Douglas Gilbert
2013-02-15 12:30 Martin Peschke
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.