> From: Richard Henderson [mailto:[email protected]] On Behalf Of Richard > Henderson > On 07/06/2015 09:26 AM, Pavel Dovgalyuk wrote: > > This patch fixes exception handling for seg_helper functions. > > > > Signed-off-by: Pavel Dovgalyuk <[email protected]> > > > No, you don't want to discriminately change every call. That was my original > point about not needing to change seg_helper.c or smm_helper.c. > > Further, any such changes would go along with the changes in translate.c to > remove the state saving there. > > I would only change those that are "normal" memory operations, like fp loads > etc. The segmentation changes are rare. The task state helpers require state > saving anyway, so requiring a TCG search is a pessimization.
I can refine the patch, but the most of the changes should remain. E.g., lcall helpers can cause an exception or not. TB ends in both cases. But icount and PC values in these two situations should be different. And lcall helpers use most of the seg functions I changed in the patch. Pavel Dovgalyuk
