linux-kernel - Re: [PATCH v2] scsi: ufs: critical health condition

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <b3f8a233-7172-41d8-a39b-49e6014a2aff@acm.org>
Date: Thu, 6 Feb 2025 10:08:41 -0800
From: Bart Van Assche <bvanassche@....org>
To: Avri Altman <Avri.Altman@....com>,
 "Martin K . Petersen" <martin.petersen@...cle.com>
Cc: "linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] scsi: ufs: critical health condition

On 2/6/25 9:47 AM, Avri Altman wrote:
>> On 2/6/25 12:54 AM, Avri Altman wrote:
>>> After some further internal discussions: The set conditions are vendor
>>> specific; The device may set it as many times it wants depending on
>>> its criticality. The spec does not define that nor what the host
>>> should do. So there is this concern that some vendors will report
>>> multiple times, while other wont. Hence, reading critical_health = 1
>>> might be misleading. What do you think?
 >
> Still not sure if you want this to be a counter?

Since the event can be reported multiple times, a counter sounds better
to me than a boolean.
>> How about emitting a uevent if a critical health condition has been reported by a
>> UFS device? See also sdev_evt_send().
> Thanks for pointing this out.
> A ufs event in enum scsi_device_event seems misplaced - looks like it was invented for unit attention codes.
> How about calling kobject_uevent() or  kobject_uevent_env() directly from the driver?

Please note that emitting a uevent is not the only possible approach for
informing user space code. Emitting a uevent is recommended if the code
that processes an event can be implemented as a shell script. Could it
be more likely that critical health events will be processed by C or C++
code rather than a shell script? If so, how about making the sysfs
attribute that reports the number of critical health events pollable? In
C and C++ code, polling a sysfs attribute requires less code than
listening for uevents.

Thanks,

Bart.