[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] correctly handle resize of empty hash tree
From: |
Christian Borntraeger |
Subject: |
Re: [Qemu-devel] [PATCH] correctly handle resize of empty hash tree |
Date: |
Thu, 25 Aug 2016 13:54:38 -0400 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.2 |
On 08/11/2016 05:41 AM, Emilio G. Cota wrote:
> On Thu, Aug 11, 2016 at 10:45:02 +0200, Igor Mammedov wrote:
>> On Wed, 10 Aug 2016 17:19:48 +0200
>> Paolo Bonzini <address@hidden> wrote:
>>> The patch makes sense, but I think we don't need to call qht_reset_size
>>> at all.
>>>
>>> tb_flush should not do anything if using KVM. There are several ways to
>>> do this:
>>>
>>> - put the tb_flush call under "if (tcg_enabled())"
>>>
>>> - add an "if (!tcg_enabled()) return;" in tb_flush
>
> I like this option the most.
I can still reproduce the crash even with 2.7.0-rc4, so the fix does not seem
to help on my system.
(gdb) bt
#0 0x000000001043a726 in qht_reset_size (ht=0x10625d98 <tcg_ctx+216>,
n_elems=32768) at /home/cborntra/REPOS/qemu/util/qht.c:422
#1 0x0000000010021ab8 in tb_flush (cpu=0x0) at
/home/cborntra/REPOS/qemu/translate-all.c:855
#2 0x0000000010076d02 in gdb_vm_state_change (opaque=0x0, running=0,
state=RUN_STATE_DEBUG) at /home/cborntra/REPOS/qemu/gdbstub.c:1276
#3 0x000000001018f106 in vm_state_notify (running=0, state=RUN_STATE_DEBUG) at
/home/cborntra/REPOS/qemu/vl.c:1585
#4 0x00000000100632e6 in do_vm_stop (state=RUN_STATE_DEBUG) at
/home/cborntra/REPOS/qemu/cpus.c:743
#5 0x0000000010065450 in vm_stop (state=RUN_STATE_DEBUG) at
/home/cborntra/REPOS/qemu/cpus.c:1476
#6 0x000000001018fc46 in main_loop_should_exit () at
/home/cborntra/REPOS/qemu/vl.c:1856
#7 0x000000001018fe6a in main_loop () at /home/cborntra/REPOS/qemu/vl.c:1912
#8 0x000000001019809c in main (argc=11, argv=0x3fffffff368,
envp=0x3fffffff3c8) at /home/cborntra/REPOS/qemu/vl.c:4604
(gdb)
Doing the "tcg_enabled()" thing does help, though.
Shall I send a patch? This should still make it into 2.7 I think.
>
> My patch to fix this issue was written thinking that tb_flush was
> strangely needed for gdb to work under KVM. If that's not the
> case, then let's go for the real fix [above].
>
> Thanks,
>
> E.
>