[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [RFC v2 04/11] tcg: comment on which functions have to

From: Alex Bennée
Subject: Re: [Qemu-devel] [RFC v2 04/11] tcg: comment on which functions have to be called with tb_lock held
Date: Thu, 05 May 2016 16:03:04 +0100
User-agent: mu4e 0.9.17; emacs

Sergey Fedorov <address@hidden> writes:

> On 05/04/16 18:32, Alex Bennée wrote:
>> diff --git a/exec.c b/exec.c
>> index f46e596..17f390e 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -826,6 +826,7 @@ int cpu_breakpoint_insert(CPUState *cpu, vaddr pc, int 
>> flags,
>>  {
>>      CPUBreakpoint *bp;
>> +    /* TODO: locking (RCU?) */
>>      bp = g_malloc(sizeof(*bp));
>>      bp->pc = pc;
> This comment is a little inconsistent. We should make access to
> breakpoint and watchpoint lists to be thread-safe in all the functions
> using them. So if we note this, it should be noted in all such places.
> Also, it's probably not a good idea to put such comment just above
> g_malloc() invocation, it could be a bit confusing. A bit more details
> would also be nice.

Good point.

I could really do with some tests to exercise the debugging interface. I
did some when I wrote the arm kvm GDB stuff (see
261f4d6d3e5445f887e070f047968e756c30cf06) but it is a) not plumbed in
and b) not really a stress test which is what you want to be sure your
handling is thread safe.

> Kind regards,
> Sergey

Alex Bennée

reply via email to

[Prev in Thread] Current Thread [Next in Thread]