[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CACeCKac6iHFwjgNn=PWth4OKCYDJRt1R6zJ+n+t6Qd4-zymyDg@mail.gmail.com>
Date: Tue, 31 Jan 2023 10:17:14 -0800
From: Prashant Malani <pmalani@...omium.org>
To: Tzung-Bi Shih <tzungbi@...nel.org>
Cc: linux-kernel@...r.kernel.org, chrome-platform@...ts.linux.dev,
bleung@...omium.org
Subject: Re: [PATCH] platform/chrome: cros_typec_vdm: Fix VDO copy
Hi Tzung-Bi,
On Mon, Jan 30, 2023 at 7:49 PM Tzung-Bi Shih <tzungbi@...nel.org> wrote:
>
> On Fri, Jan 13, 2023 at 06:26:26PM +0000, Prashant Malani wrote:
> > The usage of memcpy() affects the representation of the VDOs as they are
> > copied to the EC Host Command buffer. Specifically, all higher order
> > bits get dropped (for example: a VDO of 0x406 just gets copied as 0x6).
>
> memcpy() is byte-oriented; however, `vdo` is a pointer to u32.
>
> > Avoid this by explicitly copying each VDO in the array. The number of
> > VDOs generated by alternate mode drivers in their VDMs is almost always
> > just 1 (apart from the header) so this doesn't affect performance in a
> > meaningful way).
>
> Although the patch has applied, I am wondering if the following would be a
> better way to fix the issue:
>
> memcpy(&vdm_req.vdm_data[1], vdo, (cnt - 1) * sizeof(*vdo));
Yeah, that's right; I forgot that 'cnt' is "number of VDOs" and not
"number of bytes" :S
As I mentioned, in a vast majority of cases, the number of VDOs is just 1, so
I wouldn't bother and just leave this as it is (since there is no major benefit
apart from saving 1 line).
Powered by blists - more mailing lists