linux-kernel - Re: [PATCH 09/18] soc: qcom: ipa: GSI transactions

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:   Mon, 20 May 2019 09:34:14 -0700
From:   Evan Green <evgreen@...omium.org>
To:     Alex Elder <elder@...aro.org>
Cc:     Arnd Bergmann <arnd@...db.de>, David Miller <davem@...emloft.net>,
        Bjorn Andersson <bjorn.andersson@...aro.org>,
        Ilias Apalodimas <ilias.apalodimas@...aro.org>,
        syadagir@...eaurora.org, mjavid@...eaurora.org,
        Ben Chan <benchan@...gle.com>,
        Eric Caruso <ejcaruso@...gle.com>, abhishek.esse@...il.com,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 09/18] soc: qcom: ipa: GSI transactions

On Mon, May 20, 2019 at 7:44 AM Alex Elder <elder@...aro.org> wrote:
>
> On 5/20/19 9:43 AM, Arnd Bergmann wrote:
> > I have no idea how two 8-bit assignments could do that,
> > it sounds like a serious gcc bug, unless you mean two
> > 8-byte assignments, which would be within the range
> > of expected behavior. If it's actually 8-bit stores, please
> > open a bug against gcc with a minimized test case.
>
> Sorry, it's 8 *byte* assignments, not 8 bit.    -Alex

Is it important to the hardware that you're writing all 128 bits of
this in an atomic manner? My understanding is that while you may get
different behaviors using various combinations of
volatile/aligned/packed, this is way deep in the compiler's "free to
do whatever I want" territory. If the hardware's going to misbehave if
you don't get an atomic 128-bit write, then I don't think this has
been nailed down enough, since I think Clang or even a different
version of GCC is within its right to split the writes up differently.

Is filling out the TRE touching memory that the hardware is also
watching at the same time? Usually when the hardware cares about the
contents of a struct, there's a particular (smaller) field that can be
written out atomically. I remember USB having these structs that
needed to be filled out, but the hardware wouldn't actually slurp them
up until the smaller "token" part was non-zero. If the hardware is
scanning this struct, it might be safer to do it in two steps: 1)
flush out the filled out struct except for the field with the "go" bit
(which you'd have zeroed), then 2) set the field containing the "go"
bit.