On 11/03/15 14:25, Laszlo Ersek wrote: > (6) In closing: > > - The LZMA decompression slowdown has been quirked in the upstream > kernel. A KVM ioctl() has been introduced that allows QEMU > userspace to *disable* any quirks, but as far as I can see in > QEMU, this is not being done automatically, nor is it exposed to > the user. Therefore we can conclude that this issue has been > solved for the mid term at least.
I consider this "done" for now. > > - I have proved that the AP startup slowdown is a separate issue, > one that is caused (or "exposed") by the "other" part of host > kernel commit b18d5431acc7 -- the part that has lived on to v4.3 > un-quirked. > > I'll await Janusz's test results, and then I'll submit my proposed > fix (= extending the quirk) to the KVM mailing list. With that we > should consider this issue also resolved for OVMF, for the mid > term. Janusz reported back (thanks for that) with positive results (thanks for *that*), and I have since posted the KVM patch. (As I indicated there, that patch is not related to the SMM work.) > - Rather than auditing the MP services implementation in > UefiCpuPkg/CpuDxe *right now*, can we *PLEASE* focus on the SMM > series? > > I've been working on it (with some intermittent delays due to > external pressure) since *April*. We are so close now. This is > what remains to be done: > > - Agreement between Paolo, Jordan and Mike about implementing > broadcast SMIs. I am willing to code up whatever design is > agreed upon. Can everyone involved please prioritize this > discussion a little? We can postpone this question for now. I've withdrawn my proof of concept QEMU patch on qemu-devel, and I shelved the OVMF-side patch too. This has been made possible by Paolo's idea to return to the relaxed mode BSP/AP sync mode, *and* to mitigate the perf hit when an AP enters SMM first / only, and calls in the BSP, by shortening the relevant timeout. I just wrote a patch for the second part, and retested this approach; it works well. > - Solving the MP-related ExitBootServices() handler bug in CpuDxe. > Patches have been on the list for almost a week. While I've been > breaking my back testing and reviewing patches for others (not > just on edk2-devel, mind you), nobody has batted an eye about > that series. > > [edk2] [PATCH v2 0/4] UefiCpuPkg/CpuDxe: Fix ExitBootServices() > callback in the presence of SMIs > http://thread.gmane.org/gmane.comp.bios.edk2.devel/3518 > > Please review it, so that I can include it at the front of my > upcoming v4 SMM series. As I indicated in that thread a few hours (?) ago, I'm withdrawing this series as well, for now. The series aims to fix a bug in the ExitBootServices() handler that is only triggered by broadcast SMIs, and by returning to the relaxed mode / unicast SMI approach (see above), this bug ceases to block the SMM work. Postponed. > > - Thirty (30) patches from the SMM series still need reviews. Once > all of the above is covered, I'll update the OvmfPkg/README > patch about the status, and post version 4. I wouldn't mind if > we could commit the series still in 2015, but I'm getting less > and less hopeful. I'm in the process of formatting and double checking version 4. I'll post it later tonight. The main things to keep in mind about v4 are: - KVM + X64 + MP + S3 works (the most important use case); where S3 resume requires the Ia32X64 build - The series has no pending dependencies. All prereq patches are present in upstream edk2, QEMU, and (queued) for KVM. The cross-component diffstat is edk2/OvmfPkg/ only. Dependent on review feedback, I'll propose v4 for merging. Thanks Laszlo _______________________________________________ edk2-devel mailing list edk2-devel@lists.01.org https://lists.01.org/mailman/listinfo/edk2-devel
