linux-kernel - Re: staging: vt6656: Use common error handling code in vnt_alloc

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: <10006e33-ed90-74ec-4d59-4074caf4ecd4@web.de>
Date:   Fri, 23 Aug 2019 19:24:07 +0200
From:   Markus Elfring <Markus.Elfring@....de>
To:     Quentin Deslandes <quentin.deslandes@...ev.co.uk>
Cc:     "devel@...verdev.osuosl.org" <devel@...verdev.osuosl.org>,
        Forest Bond <forest@...ttletooquiet.net>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        LKML <linux-kernel@...r.kernel.org>,
        kernel-janitors@...r.kernel.org
Subject: Re: staging: vt6656: Use common error handling code in
 vnt_alloc_bufs()

 Your patch remove redundant code, which is fine.

Thanks for your constructive feedback.


> code you changed was simple enough to be understand quickly. I think replacing
> it with a crossed goto (even if it remove redundant code) might not be the best
> option.
>
> A solution might be to move the second loop to the top of the function

I am unsure about the relevance of the loop ordering for the affected
resource allocations.


> so you should be able to replace the end of the cleanup calls with:
>
> enomem:
> 	ret = -ENOMEM;
> free_rx:
> 	vnt_free_rx_bufs(priv);
> 	return ret;

The exception handling can eventually adjusted another bit according
to your refactoring.


> This way, only a failed call to vnt_submit_rx_urb() should jump to free_rx,

It seems that a goto statement will still be needed in an error code
part by both discussed variants.


> another failed call should jump to enomem or previously defined
> label, so we can correctly forward errors.

This view sounds promising.


> With such solution it might be worth adding a comment to describe
> that all error should be ENOMEM except for vnt_submit_rx_urb().

Can this function implementation become clearer also without
such a comment?

Regards,
Markus