- User Since
- Dec 7 2018, 3:31 PM (97 w, 4 d)
Thu, Oct 8
Mon, Sep 21
Fix "random process" crash due to leaked FPSCR bits, code cleanup and updated test plan example.
Sep 18 2020
@bdragon thank you for pointing out the 32 bit build issue!
remove extra new line
cast to u_long fixing build on 32 bit
Sep 4 2020
Code refactored, FPE exception is now identified without instruction emulate
Cleared ssr1 bits FE0 and FE1 makes user signal handler work instead of trap with SIGILL , but I don't think it's definitive solution.
added frame->srr0, frame->srr1 and frame->fixreg to output
Aug 7 2020
code refactory (still in progress)
Aug 6 2020
Jul 28 2020
Jul 24 2020
Jul 8 2020
Jul 6 2020
Remove call to internal API _kvm_err after reviewer's comment
use kvm_open instead of kvm_open2
Tested on PowerBook G4 (A1106), and it fixes kernel built with base llvm10. Nice job!
Jun 18 2020
Updated after jhibbits comments.
I don't think this patch is ready. Test shows signal being delivered to application, but I cannot catch/handle it. Need more tests.
I'm going to continue with revision and tests. Thank you @renato.riolino_eldorado.org.br for the work investigating it and providing the patch!
Jun 12 2020
Apr 29 2020
Looks good! Applied your patch and I reach the kernel. I'm using QEMU 4.2 on POWER8 machine.
Apr 15 2020
Mar 26 2020
Mar 25 2020
Mar 24 2020
Looks good to me
Mar 20 2020
Mar 16 2020
Mar 13 2020
Hi @ngie are you ok with the new "kvm_read_test" test?
I tested it on powerpc64 and amd64
Mar 12 2020
Adresses reviewers suggestions.
Mar 10 2020
Mar 9 2020
Mar 6 2020
Looks good to me. Nice job fixing it in powerpc64!
Mar 5 2020
I tested this patch with D23958/r358672 reverted. This solution makes optimized memcpy/bcopy code viable again on POWER7. It slows down some cases but it's a gain in overal, I think.
Mar 4 2020
Added code comments
System was running under load with latest change for almost 13 hours. Everything seems good.
replaced pmap_extract by pmap_kextract
I tried this patch on Powerbook5,6 with hwsensor-params-version=00 00 00 01 and didn't notice difference. With or without the patch I see the following behavior:
Mar 2 2020
Feb 28 2020
updated to match behavior of amd64/arm64 when handling an error.
Feb 27 2020
Feb 21 2020
test case improvement moved to D23783
Feb 20 2020
Feb 7 2020
Feb 3 2020
Jan 30 2020
nice job @afscoelho_gmail.com !
I just tried your patch in a image uploaded to a POWER8 instance at https://openpower.ic.unicamp.br/minicloud/ and it worked out of the box.
Please check my comment on the diff.
Jan 21 2020
Updated patch to latest HEAD (thanks to @bdragon)
Jan 17 2020
Jan 8 2020
Dec 17 2019
add more comments
Back to review as no changes planned
Dec 16 2019
enable man pages as suggested by @bdragon
build virtio modules on all powerpc platforms
Dec 14 2019
Simplified the code inspired by @imp inputs.
Dec 13 2019
Removed OS check and fixed comment
reverted PGO part
Dec 12 2019
remove explicit GCC dependency and add "modern compiler dependency" instead (compiler:c11)
rebased to current HEAD
check compiler type