Re: [PATCH] ethernet: mlx4: Delete an error message for a failed memory allocation in five functions

[Leon Romanovsky]

On Wed, Jan 03, 2018 at 01:24:59PM +0200, Tariq Toukan wrote:
>
>
> On 03/01/2018 10:06 AM, Julia Lawall wrote:
> >
> >
> > On Wed, 3 Jan 2018, Tariq Toukan wrote:
> >
> > >
> > >
> > > On 01/01/2018 10:46 PM, SF Markus Elfring wrote:
> > > > From: Markus Elfring <elfring@xxxxxxxxxxxxxxxxxxxxx>
> > > > Date: Mon, 1 Jan 2018 21:42:27 +0100
> > > >
> > > > Omit an extra message for a memory allocation failure in these functions.
> > > >
> > > > This issue was detected by using the Coccinelle software.
> > > >
> > > > Signed-off-by: Markus Elfring <elfring@xxxxxxxxxxxxxxxxxxxxx>
> > > > ---
> > >
> > > Is this an issue? Why? What is your motivation?
> > > These are error messages, very informative, appear only upon errors, and in
> > > control flow.
> >
> > Strings take up space.  Since there is a backtrace on an out of memory
> > problem, if the string does not provide any more information than the
> > position of the call, then there is not much added value.  I don't know
> > what was the string in this case.  If it provides some additional
> > information, then it would be reasonable to keep it.
>
> I don't really accept this claim...
>
> Short informative strings worth the tiny space they consume. It helps the
> users of our driver understand what went wrong in simple words, without the
> need to understand the role of the functions/callstack or being familiar
> with different parts of the driver code.
>
> In addition, some out-of-memory errors are recoverable, even though their
> backtrace is also printed. For example, in function mlx4_en_create_cq
> (appears in patch) we have a first allocation attempt (kzalloc_node) and a
> fallback (kzalloc). I'd prefer to state a clear error message only when both
> have failed, because otherwise the user might be confused whether the
> backtrace should indicate a malfunctioning interface, or not.

Tariq,

There is standard way to handle fallback in allocation and it is to
use __GFP_NOWARN flag in first allocation. So actually you pointed to the
"better-to-be-improved" function call.

Thanks

>
> Tariq
>
> >
> > julia
> >
> > > --
> > > To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
> > > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > >
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Attachment: signature.asc
Description: PGP signature