[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YfmedCAn8pK//I2R@google.com>
Date: Tue, 1 Feb 2022 20:56:20 +0000
From: Sean Christopherson <seanjc@...gle.com>
To: Nick Desaulniers <ndesaulniers@...gle.com>
Cc: Paolo Bonzini <pbonzini@...hat.com>,
Nathan Chancellor <nathan@...nel.org>,
Vitaly Kuznetsov <vkuznets@...hat.com>,
Wanpeng Li <wanpengli@...cent.com>,
Jim Mattson <jmattson@...gle.com>,
Joerg Roedel <joro@...tes.org>, kvm@...r.kernel.org,
llvm@...ts.linux.dev, linux-kernel@...r.kernel.org,
Peter Zijlstra <peterz@...radead.org>,
syzbot+6cde2282daa792c49ab8@...kaller.appspotmail.com
Subject: Re: [PATCH 1/5] Kconfig: Add option for asm goto w/ tied outputs to
workaround clang-13 bug
On Tue, Feb 01, 2022, Nick Desaulniers wrote:
> On Mon, Jan 31, 2022 at 5:08 PM Sean Christopherson <seanjc@...gle.com> wrote:
> >
> > Add a config option to guard (future) usage of asm_volatile_goto() that
> > includes "tied outputs", i.e. "+" constraints that specify both an input
> > and output parameter. clang-13 has a bug[1] that causes compilation of
> > such inline asm to fail, and KVM wants to use a "+m" constraint to
> > implement a uaccess form of CMPXCHG[2]. E.g. the test code fails with
> >
> > <stdin>:1:29: error: invalid operand in inline asm: '.long (${1:l}) - .'
> > int foo(int *x) { asm goto (".long (%l[bar]) - .\n": "+m"(*x) ::: bar); return *x; bar: return 0; }
> > ^
> > <stdin>:1:29: error: unknown token in expression
> > <inline asm>:1:9: note: instantiated into assembly here
> > .long () - .
> > ^
> > 2 errors generated.
> >
> > on clang-13, but passes on gcc (with appropriate asm goto support). The
> > bug is fixed in clang-14, but won't be backported to clang-13 as the
> > changes are too invasive/risky.
>
> LGTM.
> Reviewed-by: Nick Desaulniers <ndesaulniers@...gle.com>
>
> If you're going to respin the series, consider adding a comment in the
> source along the lines of:
> ```
> clang-14 and gcc-11 fixed this.
> ```
> or w/e. This helps us find (via grep) and remove cruft when the
> minimum supported compiler versions are updated.
Will do, a new version is definitely needed.
> Note: gcc-10 had a bug with the symbolic references to labels when
> using tied constraints.
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98096
>
> Both compilers had bugs here, and it may be worth mentioning that in
> the commit message.
Is this wording accurate?
gcc also had a similar bug[3], fixed in gcc-11, where gcc failed to
account for its behavior of assigning two numbers to tied outputs (one
for input, one for output) when evaluating symbolic references.
Powered by blists - more mailing lists