On Mon, Aug 07, 2023 at 10:27:13PM -0700, John Stultz wrote:
> On Thu, Jul 6, 2023 at 3:28 AM Richard W.M. Jones <[email protected]> wrote:
> >
> > On Thu, Jul 06, 2023 at 07:30:50AM +0100, Richard Henderson wrote:
> > > On 7/5/23 22:50, Richard W.M. Jones wrote:
> > > >tb_invalidate_phys_range_fast() *is* called, and we end up calling
> > > > tb_invalidate_phys_page_range__locked ->
> > > > tb_phys_invalidate__locked ->
> > > > do_tb_phys_invalidate
> > > >
> > > >Nevertheless the old TB (containing the call to the int3 helper) is
> > > >still called after the code has been replaced with a NOP.
> > > >
> > > >Of course there are 4 MTTCG threads so maybe another thread is in the
> > > >middle of executing the same TB when it gets invalidated.
> > >
> > > Yes.
> > >
> > > >tb_invalidate_phys_page_range__locked goes to some effort to check if
> > > >the current TB is being invalidated and restart the TB, but as far as
> > > >I can see the test can only work for the current core, and won't
> > > >restart the TB on other cores.
> > >
> > > Yes.
> > >
> > > The assumption with any of these sorts of races is that it is "as
> > > if" the other thread has already passed the location of the write
> > > within that block. But by the time this thread has finished
> > > do_tb_phys_invalidate, any other thread cannot execute the same
> > > block *again*.
> > >
> > > There's a race here, and now that I think about it, there's been mail about it in the past:
> > >
> > > https://lore.kernel.org/qemu-devel/[email protected]/
> > >
> > > We take care of the same race for user-only in translator_access, by
> > > ensuring that each translated page is read-only *before* performing
> > > the read for translation. But for system mode we grab the page
> > > locks *after* the reads. Which means there's a race.
> > >
> > > The email above describes the race pretty clearly, with a new TB
> > > being generated before the write is even complete.
> > >
> > > It'll be non-trivial fixing this, because not only do we need to
> > > grab the lock earlier, there are ordering issues for a TB that spans
> > > two pages, in that one must grab the two locks in the correct order
> > > lest we deadlock.
> >
> > Yes I can see how this is hard to fix. Even if we just lock the page
> > containing the first instruction (which we know) before doing
> > translation, we still have a problem when entering tb_link_page()
> > where we would need to only lock the second page, which might cause
> > ordering issues.
> >
> > How about a new per-page lock, which would be grabbed by
> > do_tb_phys_invalidate() and tb_gen_code(), just on the first
> > instruction? It would mean, I think, that no page can be having TBs
> > invalidated and generated at the same time.
> >
> > Or something like scanning the bytes as they are being translated,
> > generate a secure-ish checksum, then recheck it after translation and
> > discard the TB if the code changed.
>
> Hey all,
> So I've occasionally tripped over something similar in my stress
> testing of the proxy-execution patch series.
>
> This is only triggered by booting w/ qemu (7.2.2) without kvm. It's
> rare but seems to trip in bursts.
>
> The errors usually look something like:
>
> [ 21.264283] clocksource: Switched to clocksource hpet
> [ 21.273213] hrtimer: interrupt took 2400410 ns
> [ 21.369331] VFS: Disk quotas dquot_6.6.0
> [ 21.375425] int3: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC NOPTI
> [ 21.375453] CPU: 12 PID: 0 Comm: swapper/12 Not tainted
> 6.4.0-rc5-00032-g8d3f70560882-dirty #1712
> [ 21.375453] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
> BIOS 1.16.2-debian-1.16.2-1 04/01/2014
> [ 21.375453] RIP: 0010:hrtimer_start_range_ns+0x1ab/0x3d0
> [ 21.375453] Code: 0f 85 70 ff ff ff 44 8b 7c 24 18 4c 63 65 08 4c
> 89 f3 e8 c8 86 0c 01 89 c0 41 83 e7 02 48 03 1c c5 40 0c ac b1 48 89
> 1c 24 0f <1f> 44 00 00 45 85 ff 0f 84 e3 01 00 00 4d 8d 7c 24 01 49 c1
> e4 07
> [ 21.375453] RSP: 0018:ffffbf4b000f7e60 EFLAGS: 00000086
> [ 21.375453] RAX: 000000000000000c RBX: ffff984ff85239c0 RCX: 00000004ef14ffc0
> [ 21.375453] RDX: 7fffffffffffffff RSI: ffffffffb1973381 RDI: ffffffffb1976d48
> [ 21.375453] RBP: ffff984ff8523a40 R08: 0000000000000001 R09: 0000000000000000
> [ 21.375453] R10: 00000000e8003900 R11: 0000000005673366 R12: 0000000000000000
> [ 21.375453] R13: ffff984ff85242f0 R14: 00000000000239c0 R15: 0000000000000002
> [ 21.375453] FS: 0000000000000000(0000) GS:ffff984ff8500000(0000)
> knlGS:0000000000000000
> [ 21.375453] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 21.375453] CR2: 0000000000000000 CR3: 0000000045e62000 CR4: 00000000000006e0
> [ 21.375453] Call Trace:
> [ 21.375453] <TASK>
> [ 21.375453] ? die+0x2d/0x80
> [ 21.375453] ? exc_int3+0xf3/0x100
> [ 21.375453] ? asm_exc_int3+0x35/0x40
> [ 21.375453] ? hrtimer_start_range_ns+0x1ab/0x3d0
> [ 21.375453] ? hrtimer_start_range_ns+0x1ab/0x3d0
> [ 21.375453] tick_nohz_restart+0x72/0x90
> [ 21.375453] tick_nohz_idle_exit+0xc7/0xf0
> [ 21.375453] do_idle+0x160/0x220
> [ 21.375453] cpu_startup_entry+0x14/0x20
> [ 21.375453] start_secondary+0xf5/0x100
> [ 21.375453] secondary_startup_64_no_verify+0x10b/0x10b
> [ 21.375453] </TASK>
>
>
> Curious if anyone has found a workaround?
Yes, it should be fixed upstream. You will need these two commits:
commit deba78709ae8ce103e2248413857747f804cd1ef
Author: Richard Henderson <[email protected]>
Date: Thu Jul 6 17:55:48 2023 +0100
accel/tcg: Always lock pages before translation
commit cb62bd15e14e304617d250158b77d0deb032f032
Author: Richard Henderson <[email protected]>
Date: Thu Jul 6 08:45:13 2023 +0100
accel/tcg: Split out cpu_exec_longjmp_cleanup
Rich.
--
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-top is 'top' for virtual machines. Tiny program with many
powerful monitoring features, net stats, disk stats, logging, etc.
http://people.redhat.com/~rjones/virt-top