On Sat, Aug 12, 2017 at 4:42 AM, Edgar E. Iglesias <edgar.igles...@xilinx.com> wrote: > On Sat, Aug 12, 2017 at 11:24:30AM +0100, Peter Maydell wrote: >> On 11 August 2017 at 23:17, Alistair Francis >> <alistair.fran...@xilinx.com> wrote: >> > I found some issues with the way exclusive store was working. This patch >> > series seems to fix the test cases that were failing for me. >> > >> > The first patch is just a simple adjustment. >> > >> > The third patch fixes the main bug I was seeing. >> > >> > The second patch is left over from the RFC that seems like it is still a >> > good idea. >> > >> > Changes from RFC: >> > - Rewrite the third patch to correctly fix the issue. >> > >> > Alistair Francis (3): >> > target/arm: Update the memops for exclusive load >> > tcg/tcg-op: Expose the tcg_gen_ext_i* functions >> > target/arm: Correct exclusive store cmpxchg memop mask >> > >> > target/arm/translate-a64.c | 4 ++-- >> > tcg/tcg-op.c | 4 ++-- >> > tcg/tcg-op.h | 2 ++ >> > 3 files changed, 6 insertions(+), 4 deletions(-) >> >> Is this series (or at least patches 1 and 3) worth putting >> into 2.10 ? > > > I would vote for including it...
The only reason not to is because this bug has been in QEMU for a while, so it obviously isn't hit very often. In saying that it is a pretty big issue (32-bit pair stores are completely broken) which might become an issue during the 2.10 support window and I don't see many complications from including the series. I agree with Edgar, probably worth including. Thanks, Alistair > > Cheers, > Edgar
