linux-kernel - Re: [PATCHv2] usb: gadget: storage: optional SCSI WRITE FUA bit

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]

Date:	Wed, 14 Jul 2010 18:05:07 +0300
From:	Andy Shevchenko <andy.shevchenko@...il.com>
To:	Michał Nazarewicz <m.nazarewicz@...sung.com>
Cc:	linux-kernel@...r.kernel.org,
	Andy Shevchenko <ext-andriy.shevchenko@...ia.com>,
	Denis Karpov <ext-denis.2.karpov@...ia.com>,
	Adrian Hunter <adrian.hunter@...ia.com>,
	Alan Stern <stern@...land.harvard.edu>,
	David Brownell <dbrownell@...rs.sourceforge.net>,
	Greg Kroah-Hartman <gregkh@...e.de>, linux-usb@...r.kernel.org
Subject: Re: [PATCHv2] usb: gadget: storage: optional SCSI WRITE FUA bit

2010/7/14 Michał Nazarewicz <m.nazarewicz@...sung.com>:
> On Wed, 14 Jul 2010 15:44:58 +0200, Andy Shevchenko
> <andy.shevchenko@...il.com> wrote:
>>>> + *     fua=b[,b...]            Default false, booleans for ignore FUA
>>>> flag
>>>> + *                                     in SCSI WRITE(6,10,12) commands
>>>
>>> I wonder if it makes sense to make it per-LUN.  I would imagine
>>> that it's great to ignore FUA if the device has its own power supply
>>> in which case after disconnect the data won't be lost.  This is a
>>> per-device property not really per-LUN.  As such I'd make this option
>>> global for the gadget.
>
>> Make sense only for removable media with one partition.
>> Otherwise. why we have sync option per partition f.e., not per device?
>
> Ah, OK, I see why this is per LUN.  You want to be able not to ignore
> FUA if the backing storage is a removable media, right?
In instance, or vise versa.
So, the user could decide if he wants to avoid this flag for one LUN
or for another.

>>>> +       if (sscanf(buf, "%d", &i) != 1)
>>>> +               return -EINVAL;
>>> Why not simple_strtol() directly?
>> I did it in the same way like fsg_store_ro() does.
>> I have no objections to back to previous solution.
> OK.  I'd use simpre_strol() myself.  Maybe even patched fsg_store_ro().
Agree, but better to do series of patches then, I guess.

>>>> +
>>>> +       if (curlun->fua)
>>>> +               fsg_lun_fsync_sub(curlun);
>
>>> Shouldn't that read something like:
>>>
>>> +       if (!curlun->fua && i)
>>> +               fsg_lun_fsync_sub(curlun);
>>>
>>> ie. there is no sense in syncing if FUA was active (in which case all
>>> writes were synced already, right?) or if the new value is false (since
>>> then user does not won't syncing).
>
>> The idea is to sync data before switching from async mode.
>
> But there can be a case of switching from async to async when syncing
> is not necessary.  That's why I proposed the &&.  With fua = 1 meaning
> ignore the flag my proposal would be:
>
> +       if (!i && curlun->fua)
> +               fsg_lun_fsync_sub(curlun);
Makes sense.

>> Actually fua = 1 means ignorance of that flag.
> ignore_fua would be better name then I think.  This also stands for
> module parameter.
I already thought about. Rather I agree with you.


-- 
With Best Regards,
Andy Shevchenko
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/