[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <b03a25f0-ea67-3e1b-ffc7-3d9b9f6e838e@redhat.com>
Date: Mon, 4 Feb 2019 06:58:49 +0100
From: Laura Abbott <labbott@...hat.com>
To: David Disseldorp <ddiss@...e.de>
Cc: Christoph Hellwig <hch@....de>,
"Martin K. Petersen" <martin.petersen@...cle.com>,
"Nicholas A. Bellinger" <nab@...ux-iscsi.org>,
linux-scsi@...r.kernel.org, target-devel@...r.kernel.org,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Lee Duncan <lduncan@...e.com>
Subject: Re: Userspace regression with 6baca7601bde ("scsi: target: drop
unused pi_prot_format attribute storage")
On 2/4/19 1:40 AM, David Disseldorp wrote:
> Hi Laura,
>
> Thanks for the report...
>
> On Sun, 3 Feb 2019 17:56:00 +0100, Laura Abbott wrote:
>
>> Fedora got a bug report of a new permission denied error with 5.0-rc2:
>>
>>> File "/usr/lib/python3.7/site-packages/rtslib_fb/utils.py", line 100, in fread
>>> with open(path, 'r') as file_fd:
>>> PermissionError: [Errno 13] Permission denied: '/sys/kernel/config/target/core/fileio_28/xxx/attrib/pi_prot_format'
>>
>> This looks like an intentional behavior change with
>>
>> commit 6baca7601bdee2e57f20c45d63eb53b89b33e816
>> Author: David Disseldorp <ddiss@...e.de>
>> Date: Fri Nov 23 18:36:11 2018 +0100
>>
>> scsi: target: drop unused pi_prot_format attribute storage
>>
>> On write, the pi_prot_format configfs attribute invokes the device
>> format_prot() callback if present. Read dumps the contents of
>> se_dev_attrib.pi_prot_format which is always zero. Make the configfs
>> attribute write-only, and drop the always zero se_dev_attrib.pi_prot_format
>> storage.
>>
>> Signed-off-by: David Disseldorp <ddiss@...e.de>
>> Reviewed-by: Christoph Hellwig <hch@....de>
>> Signed-off-by: Martin K. Petersen <martin.petersen@...cle.com>
>>
>>
>> Unfortunately, existing code that's opening with read permissions is now broken.
>> Can this be reverted? Full bug at https://bugzilla.redhat.com/show_bug.cgi?id=1667505
>
> Lee (cc'ed) pinged me a couple of days ago about the same issue.
> My preference would be to add back a dummy read handler without the
> corresponding (unused) se_dev_attrib.pi_prot_format member.
> I'll prepare something tomorrow with this, but if it's urgent then I'd
> also be okay with a straight revert.
>
> Cheers, David
>
A fix is fine by me. Thanks for the prompt response.
Thanks,
Laura
Powered by blists - more mailing lists