From: Bill Wendling <morbo@google.com>
To: Thomas Huth <thuth@redhat.com>
Cc: kvm list <kvm@vger.kernel.org>,
Paolo Bonzini <pbonzini@redhat.com>,
alexandru.elisei@arm.com, Jim Mattson <jmattson@google.com>
Subject: Re: [kvm-unit-tests PATCH 1/1] x86: use pointer for end of exception table
Date: Tue, 15 Oct 2019 01:06:40 -0700 [thread overview]
Message-ID: <CAGG=3QVzF+Obz4ONuV-GPwz8v-St0VbkbDf2J9S3dS3o09pNzg@mail.gmail.com> (raw)
In-Reply-To: <6656e70c-4866-55e8-108b-9bbb2c6fd081@redhat.com>
On Tue, Oct 15, 2019 at 12:41 AM Thomas Huth <thuth@redhat.com> wrote:
>
> On 13/10/2019 09.18, Bill Wendling wrote:
> > Two global objects can't have the same address in C. Clang uses this
> > fact to omit the check on the first iteration of the loop in
> > check_exception_table.
> >
> > Signed-off-by: Bill Wendling <morbo@google.com>
> > ---
> > lib/x86/desc.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/lib/x86/desc.c b/lib/x86/desc.c
> > index 451f504..cfc449f 100644
> > --- a/lib/x86/desc.c
> > +++ b/lib/x86/desc.c
> > @@ -41,7 +41,7 @@ struct ex_record {
> > unsigned long handler;
> > };
> >
> > -extern struct ex_record exception_table_start, exception_table_end;
> > +extern struct ex_record exception_table_start, *exception_table_end;
> >
> > static const char* exception_mnemonic(int vector)
> > {
> > @@ -113,7 +113,7 @@ static void check_exception_table(struct ex_regs *regs)
> > (((regs->rflags >> 16) & 1) << 8);
> > asm("mov %0, %%gs:4" : : "r"(ex_val));
> >
> > - for (ex = &exception_table_start; ex != &exception_table_end; ++ex) {
> > + for (ex = &exception_table_start; ex != (void*)&exception_table_end; ++ex) {
> > if (ex->rip == regs->rip) {
> > regs->rip = ex->handler;
> > return;
> >
>
> That looks like quite an ugly hack to me - and if clang gets a little
> bit smarter, I'm pretty sure it will optimize this away, too.
>
I agree. I was thinking about this actually and this patch looks like
it'll work. I'll generate a new version of the patches.
diff --git a/lib/x86/desc.c b/lib/x86/desc.c
index 451f504..4002203 100644
--- a/lib/x86/desc.c
+++ b/lib/x86/desc.c
@@ -113,7 +113,7 @@ static void check_exception_table(struct ex_regs *regs)
(((regs->rflags >> 16) & 1) << 8);
asm("mov %0, %%gs:4" : : "r"(ex_val));
- for (ex = &exception_table_start; ex != &exception_table_end; ++ex) {
+ for (ex = &exception_table_start; ex < &exception_table_end; ++ex) {
if (ex->rip == regs->rip) {
regs->rip = ex->handler;
return;
> Could you please check if something like this works, too:
>
> diff --git a/lib/x86/desc.c b/lib/x86/desc.c
> index 451f504..0e9779b 100644
> --- a/lib/x86/desc.c
> +++ b/lib/x86/desc.c
> @@ -41,7 +41,8 @@ struct ex_record {
> unsigned long handler;
> };
>
> -extern struct ex_record exception_table_start, exception_table_end;
> +extern struct ex_record exception_table_start;
> +extern long exception_table_size;
>
> static const char* exception_mnemonic(int vector)
> {
> @@ -108,12 +109,13 @@ static void check_exception_table(struct ex_regs
> *regs)
> {
> struct ex_record *ex;
> unsigned ex_val;
> + int cnt = exception_table_size / sizeof(struct ex_record);
>
> ex_val = regs->vector | (regs->error_code << 16) |
> (((regs->rflags >> 16) & 1) << 8);
> asm("mov %0, %%gs:4" : : "r"(ex_val));
>
> - for (ex = &exception_table_start; ex != &exception_table_end; ++ex) {
> + for (ex = &exception_table_start; cnt-- > 0; ++ex) {
> if (ex->rip == regs->rip) {
> regs->rip = ex->handler;
> return;
> diff --git a/x86/flat.lds b/x86/flat.lds
> index a278b56..108fad5 100644
> --- a/x86/flat.lds
> +++ b/x86/flat.lds
> @@ -9,6 +9,7 @@ SECTIONS
> exception_table_start = .;
> *(.data.ex)
> exception_table_end = .;
> + exception_table_size = exception_table_end -
> exception_table_start;
> }
> . = ALIGN(16);
> .rodata : { *(.rodata) }
>
> Thomas
next prev parent reply other threads:[~2019-10-15 8:06 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-12 7:44 [kvm-unit-tests PATCH 1/2] Use a status enum for reporting pass/fail Bill Wendling
2019-10-12 7:44 ` [kvm-unit-tests PATCH 2/2] x86: use pointer for end of exception table Bill Wendling
2019-10-13 7:18 ` [kvm-unit-tests PATCH 1/1] " Bill Wendling
2019-10-15 7:41 ` Thomas Huth
2019-10-15 8:06 ` Bill Wendling [this message]
2019-10-12 8:20 ` [kvm-unit-tests PATCH 2/2] Use a status enum for reporting pass/fail Bill Wendling
2019-10-14 16:26 ` [kvm-unit-tests PATCH 1/2] " Sean Christopherson
2019-10-14 18:23 ` Bill Wendling
2019-10-15 20:46 ` [kvm-unit-tests v2 PATCH 0/2] Clang compilation fixes Bill Wendling
2019-10-15 20:46 ` [kvm-unit-tests v2 PATCH 1/2] lib: use a status enum for reporting pass/fail Bill Wendling
2019-10-15 20:46 ` [kvm-unit-tests v2 PATCH 2/2] x86: don't compare two global objects' addrs for inequality Bill Wendling
2019-10-16 5:53 ` Thomas Huth
2019-10-21 15:33 ` Paolo Bonzini
2019-10-21 15:32 ` [kvm-unit-tests PATCH 1/2] Use a status enum for reporting pass/fail Paolo Bonzini
2019-10-29 18:51 ` Bill Wendling
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAGG=3QVzF+Obz4ONuV-GPwz8v-St0VbkbDf2J9S3dS3o09pNzg@mail.gmail.com' \
--to=morbo@google.com \
--cc=alexandru.elisei@arm.com \
--cc=jmattson@google.com \
--cc=kvm@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=thuth@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).