- 08 Apr, 2012 40 commits
-
-
Alexander Graf authored
When we get a performance monitor interrupt, we need to make sure that the host receives it. So reinject it like we reinject the other host destined interrupts. Signed-off-by:
Alexander Graf <agraf@suse.de> Signed-off-by:
Avi Kivity <avi@redhat.com>
-
Alexander Graf authored
When reinjecting an interrupt into the host interrupt handler after we're back in host kernel land, we need to tell the kernel where the interrupt happened. We can't tell it that we were in guest state, because that might lead to random code walking host addresses. So instead, we tell it that we came from the interrupt reinject code. This helps getting reasonable numbers out of perf. Signed-off-by:
-
Alexander Graf authored
When during guest context we get a performance monitor interrupt, we currently bail out and oops. Let's route it to its correct handler instead. Signed-off-by:
-
Alexander Graf authored
The tlbncfg registers should be populated with their respective TLB's values. Fix the obvious typo. Signed-off-by:
-
Alexander Graf authored
There was some unused code in the exit code path that must have been a leftover from earlier iterations. While it did no harm, it's superfluous and thus should be removed. Signed-off-by:
-
Alexander Graf authored
The comment for program interrupts triggered when using bookehv was misleading. Update it to mention why MSR_GS indicates that we have to inject an interrupt into the guest again, not emulate it. Signed-off-by:
-
Alexander Graf authored
When during guest execution we get a machine check interrupt, we don't know how to handle it yet. So let's add the error printing code back again that we dropped accidently earlier and tell user space that something went really wrong. Signed-off-by:
-
Alexander Graf authored
For BookE HV the guest visible MSR is shared->msr and is identical to the MSR that is in use while the guest is running, because we can't trap reads from/to MSR. So shadow_msr is unused there. Indicate that with a comment. Signed-off-by:
-
Alexander Graf authored
We need to make sure that no MAS updates happen automatically while we have the guest MAS registers loaded. So move the disabling code a bit higher up so that it covers the full time we have guest values in MAS registers. The race this patch fixes should never occur, but it makes the code a bit more logical to do it this way around. Signed-off-by:
-
Alexander Graf authored
The SET_VCPU macro is a leftover from times when the vcpu struct wasn't stored in the thread on vcpu_load/put. It's not needed anymore. Remove it. Signed-off-by:
-
Alexander Graf authored
Instead if doing #ifndef CONFIG_64BIT ... #else ... #endif we should rather do #ifdef CONFIG_64BIT ... #else ... #endif which is a lot easier to read. Change the bookehv implementation to stick with this rule. Signed-off-by:
-
Alexander Graf authored
When using exit timing stats, we clobber r9 in the NEED_EMU case, so better move that part down a few lines and fix it that way. Signed-off-by:
-
Alexander Graf authored
The semantics of BOOKE_IRQPRIO_MAX changed to denote the highest available irqprio + 1, so let's reflect that in the code too. Signed-off-by:
-
Alexander Graf authored
Instead of checking whether we should reschedule only when we exited due to an interrupt, let's always check before entering the guest back again. This gets the target more in line with the other archs. Also while at it, generalize the whole thing so that eventually we could have a single kvmppc_prepare_to_enter function for all ppc targets that does signal and reschedule checking for us. Signed-off-by:
-
Alexander Graf authored
When we fail to emulate an instruction for the guest, we better go in and tell it that we failed to emulate it, by throwing an illegal instruction exception. Please beware that we basically never get around to telling the guest that we failed thanks to the debugging code right above it. If user space however decides that it wants to ignore the debug, we would at least do "the right thing" afterwards. Signed-off-by:
-
Alexander Graf authored
The e500mc patches left some debug code in that we don't need. Remove it. Signed-off-by:
-
Alexander Graf authored
We can't run e500v2 kvm on e500mc kernels, so indicate that by making the 2 options mutually exclusive in kconfig. Signed-off-by:
-
Alexander Graf authored
The CONFIG_KVM_E500 option really indicates that we're running on a V2 machine, not on a machine of the generic E500 class. So indicate that properly and change the config name accordingly. Signed-off-by:
-
Alexander Graf authored
There's always a chance we're unable to read a guest instruction. The guest could have its TLB mapped execute-, but not readable, something odd happens and our TLB gets flushed. So it's a good idea to be prepared for that case and have a fallback that allows us to fix things up in that case. Add fixup code that keeps guest code from potentially crashing our host kernel. Signed-off-by:
-
Alexander Graf authored
If we hit any exception whatsoever in the restore path and r1/r2 aren't the host registers, we don't get a working oops. So it's always a good idea to restore them as early as possible. This time, it actually has practical reasons to do so too, since we need to have the host page fault handler fix up our guest instruction read code. And for that to work we need r1/r2 restored. Signed-off-by:
-
Alexander Graf authored
When setting MSR for an e500mc guest, we implicitly always set MSR_GS to make sure the guest is in guest state. Since we have this implicit rule there, we don't need to explicitly pass MSR_GS to set_msr(). Remove all explicit setters of MSR_GS. Signed-off-by:
-
Alexander Graf authored
When one vcpu wants to kick another, it can issue a special IPI instruction called msgsnd. This patch emulates this instruction, its clearing counterpart and the infrastructure required to actually trigger that interrupt inside a guest vcpu. With this patch, SMP guests on e500mc work. Signed-off-by:
-
Scott Wood authored
Add processor support for e500mc, using hardware virtualization support (GS-mode). Current issues include: - No support for external proxy (coreint) interrupt mode in the guest. Includes work by Ashish Kalra <Ashish.Kalra@freescale.com>, Varun Sethi <Varun.Sethi@freescale.com>, and Liu Yu <yu.liu@freescale.com>. Signed-off-by:
Scott Wood <scottwood@freescale.com> Signed-off-by:
-
Scott Wood authored
e500mc has a normal PPC FPU, rather than SPE which is found on e500v1/v2. Based on code from Liu Yu <yu.liu@freescale.com>. Signed-off-by:
-
Scott Wood authored
Chips such as e500mc that implement category E.HV in Power ISA 2.06 provide hardware virtualization features, including a new MSR mode for guest state. The guest OS can perform many operations without trapping into the hypervisor, including transitions to and from guest userspace. Since we can use SRR1[GS] to reliably tell whether an exception came from guest state, instead of messing around with IVPR, we use DO_KVM similarly to book3s. Current issues include: - Machine checks from guest state are not routed to the host handler. - The guest can cause a host oops by executing an emulated instruction in a page that lacks read permission. Existing e500/4xx support has the same problem. Includes work by Ashish Kalra <Ashish.Kalra@freescale.com>, Varun Sethi <Varun.Sethi@freescale.com>, and Liu Yu <yu.liu@freescale.com>. Signed-off-by:
-
Scott Wood authored
DO_KVM will need to identify the particular exception type. There is an existing set of arbitrary numbers that Linux passes, but it's an undocumented mess that sort of corresponds to server/classic exception vectors but not really. Signed-off-by:
-
Scott Wood authored
tlbilx is the new, preferred invalidation instruction. It is not found on e500 prior to e500mc, but there should be no harm in supporting it on all e500. Based on code from Ashish Kalra <Ashish.Kalra@freescale.com>. Signed-off-by:
-
Scott Wood authored
Rather than invalidate everything when a TLB1 entry needs to be taken down, keep track of which host TLB1 entries are used for a given guest TLB1 entry, and invalidate just those entries. Based on code from Ashish Kalra <Ashish.Kalra@freescale.com> and Liu Yu <yu.liu@freescale.com>. Signed-off-by:
-
Scott Wood authored
The PID handling is e500v1/v2-specific, and is moved to e500.c. The MMU sregs code and kvmppc_core_vcpu_translate will be shared with e500mc, and is moved from e500.c to e500_tlb.c. Partially based on patches from Liu Yu <yu.liu@freescale.com>. Signed-off-by:
-
Scott Wood authored
Move vcpu to the beginning of vcpu_e500 to give it appropriate prominence, especially if more fields end up getting added to the end of vcpu_e500 (and vcpu ends up in the middle). Remove gratuitous "extern" and add parameter names to prototypes. Signed-off-by:
-
Scott Wood authored
Keeping two separate headers for e500-specific things was a pain, and wasn't even organized along any logical boundary. There was TLB stuff in <asm/kvm_e500.h> despite the existence of arch/powerpc/kvm/e500_tlb.h, and nothing in <asm/kvm_e500.h> needed to be referenced from outside arch/powerpc/kvm. Signed-off-by:
-
Scott Wood authored
This is in preparation for merging in the contents of arch/powerpc/include/asm/kvm_e500.h. Signed-off-by:
-
Scott Wood authored
e500mc will want to do lpid allocation/deallocation here. Signed-off-by:
-
Scott Wood authored
This gives us a place to put load/put actions that correspond to code that is booke-specific but not specific to a particular core. Signed-off-by:
-
Scott Wood authored
We'll use it on e500mc as well. Signed-off-by:
-
Scott Wood authored
Split e500 (v1/v2) and e500mc/e5500 to allow optimization of feature checks that differ between the two. Signed-off-by:
-
Scott Wood authored
Currently 32-bit only cares about this for choice of exception vector, which is done in core-specific code. However, KVM will want to distinguish as well. Signed-off-by:
-
Takuya Yoshikawa authored
Now that we do neither double buffering nor heuristic selection of the write protection method these are not needed anymore. Note: some drivers have their own implementation of set_bit_le() and making it generic needs a bit of work; so we use test_and_set_bit_le() and will later replace it with generic set_bit_le(). Signed-off-by:
Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp> Signed-off-by:
-
Takuya Yoshikawa authored
We have seen some problems of the current implementation of get_dirty_log() which uses synchronize_srcu_expedited() for updating dirty bitmaps; e.g. it is noticeable that this sometimes gives us ms order of latency when we use VGA displays. Furthermore the recent discussion on the following thread "srcu: Implement call_srcu()" http://lkml.org/lkml/2012/1/31/211 also motivated us to implement get_dirty_log() without SRCU. This patch achieves this goal without sacrificing the performance of both VGA and live migration: in practice the new code is much faster than the old one unless we have too many dirty pages. Implementation: The key part of the implementation is the use of xchg() operation for clearing dirty bits atomically. Since this allows us to update only BITS_PER_LONG pages at once, we need to iterate over the dirty bitmap until every dirty bit is cleared again for the next call. Although some people may worry about the problem of using the atomic memory instruction many times to the concurrently accessible bitmap, it is usually accessed with mmu_lock held and we rarely see concurrent accesses: so what we need to care about is the pure xchg() overheads. Another point to note is that we do not use for_each_set_bit() to check which ones in each BITS_PER_LONG pages are actually dirty. Instead we simply use __ffs() in a loop. This is much faster than repeatedly call find_next_bit(). Performance: The dirty-log-perf unit test showed nice improvements, some times faster than before, except for some extreme cases; for such cases the speed of getting dirty page information is much faster than we process it in the userspace. For real workloads, both VGA and live migration, we have observed pure improvements: when the guest was reading a file during live migration, we originally saw a few ms of latency, but with the new method the latency was less than 200us. Signed-off-by: -
Takuya Yoshikawa authored
Dropped such mappings when we enabled dirty logging and we will never create new ones until we stop the logging. For this we introduce a new function which can be used to write protect a range of PT level pages: although we do not need to care about a range of pages at this point, the following patch will need this feature to optimize the write protection of many pages. Signed-off-by:
-
