On 1/29/2024 7:43 AM, Alexander Lobakin wrote:
From: Brady, Alan <alan.br...@intel.com>
Date: Mon, 29 Jan 2024 16:12:06 +0100
-----Original Message-----
From: Lobakin, Aleksander <aleksander.loba...@intel.com>
Sent: Monday, January 29, 2024 5:24 AM
To: Brady, Alan <alan.br...@intel.com>
Cc: intel-wired-...@lists.osuosl.org; net...@vger.kernel.org;
willemdebruijn.ker...@gmail.com; Bagnucki, Igor
<igor.bagnu...@intel.com>; Kitszel, Przemyslaw
<przemyslaw.kits...@intel.com>
Subject: Re: [Intel-wired-lan] [PATCH v2 0/7 iwl-next] idpf: refactor virtchnl
messages
From: Alan Brady <alan.br...@intel.com>
Date: Thu, 25 Jan 2024 21:47:40 -0800
The motivation for this series has two primary goals. We want to
enable support of multiple simultaneous messages and make the channel
more robust. The way it works right now, the driver can only send and
receive a single message at a time and if something goes really wrong,
it can lead to data corruption and strange bugs.
[...]
There are a fistful of functions in this series and IDPF's virtchnl code in
general
that allocate a memory chunk via kzalloc() family and then free it at the end of
the function, i.e. the lifetime of those buffers are the lifetime of the
function.
Since recently, we have auto-variables in the kernel, so that the pieces I
described could be converted to:
struct x *ptr __free(kfree) = NULL;
ptr = kzalloc(sizeof(*x), GPF_KERNEL);
// some code
return 0; // kfree() is not needed anymore
err:
return err; // here as well
That would allow to simplify the code and reduce its size.
I'd like you to convert such functions to use auto-variables.
Certainly, should be straightforward and make the code much better, sounds good
to me. Just to clarify I'm only going to mess with the virtchnl functions I've
otherwise altered in this patch series to maintain appropriate scope, yes?
Yes, only virtchnl functions. New functions that you introduce 100%, the
rest only if you touch them.
I've gone ahead and submitted a v3 with the suggested changes. However,
I would also like to follow this up with another series that addresses
the entire idpf driver where we can not only use the automatic variables
for kfree, but also make the various locks we have much better. I do end
up touching some of the locks I introduced here, but I need to add a
guard to include/linux/spinlock.h to convert all of it. I think it still
makes sense to not include those changes in this series to avoid
polluting it if possible, but there will be a dependency. Also thank you
for pointing this out, I think it's a massive improvement to the driver
code.
