Re: [Qemu-devel] [PATCHv2 0/4] Set eflags prior to segments; fix cpl in SMM

[Richard Henderson]

On 04/30/2014 01:37 AM, Paolo Bonzini wrote:
> Il 29/04/2014 22:37, Kevin O'Connor ha scritto:
>> This patch series is an update to the patch I sent at:
>>
>> https://lists.gnu.org/archive/html/qemu-devel/2014-04/msg04028.html
>>
>> The goal of the patch series is to fix the SMM handler so that it runs
>> with the proper cpl.  I've broken the series up based on the feedback
>> from Paolo.
>>
>> As with the previous patch, this series moves cr0 and eflags changes
>> above cpu_x86_load_seg_cache() calls as cpu_x86_load_seg_cache()
>> inspects those flags.  I didn't move the eflags setting in
>> helper_ret_protected() because I don't think I can easily move the
>> eflags logic (it's not clear to me that it is safe to change eflags
>> prior to exception checks) and helper_ret_protected() doesn't change
>> any eflags bits that cpu_x86_load_seg_cache() cares about.
>>
>> That said, continuing to have cpu_x86_load_seg_cache() depend on cr0
>> and eflags is painful.  I do wonder if it would be better to calculate
>> hflags lazily to eliminate these ordering issues.
> 
> Reviewed-by: Paolo Bonzini <address@hidden>

Two style nits, but otherwise

Reviewed-by: Richard Henderson <address@hidden>

> 
> Richard, Andreas, who is going to take this?  I can do it too since I plan to
> put my soft-reset target-i386/ patches in uq/master after they are reviewed.

Please do.


r~