- 22 May, 2011 40 commits
-
-
Takuya Yoshikawa authored
The address of the gpte was already calculated and stored in ptep_user before entering cmpxchg_gpte(). This patch makes cmpxchg_gpte() to use that to make it clear that we are using the same address during walk_addr_generic(). Note that the unlikely annotations are used to show that the conditions are something unusual rather than for performance. Signed-off-by:
Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp> Signed-off-by:
Marcelo Tosatti <mtosatti@redhat.com>
-
OGAWA Hirofumi authored
Like the following, mmu_notifier can be called after registering immediately. So, kvm have to initialize kvm->mmu_lock before it. BUG: spinlock bad magic on CPU#0, kswapd0/342 lock: ffff8800af8c4000, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0 Pid: 342, comm: kswapd0 Not tainted 2.6.39-rc5+ #1 Call Trace: [<ffffffff8118ce61>] spin_bug+0x9c/0xa3 [<ffffffff8118ce91>] do_raw_spin_lock+0x29/0x13c [<ffffffff81024923>] ? flush_tlb_others_ipi+0xaf/0xfd [<ffffffff812e22f3>] _raw_spin_lock+0x9/0xb [<ffffffffa0582325>] kvm_mmu_notifier_clear_flush_young+0x2c/0x66 [kvm] [<ffffffff810d3ff3>] __mmu_notifier_clear_flush_young+0x2b/0x57 [<ffffffff810c8761>] page_referenced_one+0x88/0xea [<ffffffff810c89bf>] page_referenced+0x1fc/0x256 [<ffffffff810b2771>] shrink_page_list+0x187/0x53a [<ffffffff810b2ed7>] shrink_inactive_list+0x1e0/0x33d [<ffffffff810acf95>] ? determine_dirtyable_memory+0x15/0x27 [<ffffffff812e90ee>] ? call_function_single_interrupt+0xe/0x20 [<ffffffff810b3356>] shrink_zone+0x322/0x3de [<ffffffff810a9587>] ? zone_watermark_ok_safe+0xe2/0xf1 [<ffffffff810b3928>] kswapd+0x516/0x818 [<ffffffff810b3412>] ? shrink_zone+0x3de/0x3de [<ffffffff81053d17>] kthread+0x7d/0x85 [<ffffffff812e9394>] kernel_thread_helper+0x4/0x10 [<ffffffff81053c9a>] ? __init_kthread_worker+0x37/0x37 [<ffffffff812e9390>] ? gs_change+0xb/0xb Signed-off-by:
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> Signed-off-by:
Avi Kivity <avi@redhat.com>
-
Pekka Enberg authored
Document KVM_CAP_NR_VCPUS that can be used by the userspace to determine maximum number of VCPUs it can create with the KVM_CREATE_VCPU ioctl. Cc: Avi Kivity <avi@redhat.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Jan Kiszka <jan.kiszka@web.de> Signed-off-by:
Pekka Enberg <penberg@kernel.org> Signed-off-by:
-
Gleb Natapov authored
KVM does not hold any references to rcu protected data when it switches CPU into a guest mode. In fact switching to a guest mode is very similar to exiting to userspase from rcu point of view. In addition CPU may stay in a guest mode for quite a long time (up to one time slice). Lets treat guest mode as quiescent state, just like we do with user-mode execution. Signed-off-by:
Gleb Natapov <gleb@redhat.com> Signed-off-by:
-
Avi Kivity authored
* commit '29ce8310': (34 commits) rcu: provide rcu_virt_note_context_switch() function. rcu: get rid of signed overflow in check_cpu_stall() rcu: optimize rcutiny rcu: prevent call_rcu() from diving into rcu core if irqs disabled rcu: further lower priority in rcu_yield() rcu: introduce kfree_rcu() rcu: fix spelling rcu: call __rcu_read_unlock() in exit_rcu for tree RCU rcu: Converge TINY_RCU expedited and normal boosting rcu: remove useless ->boosted_this_gp field rcu: code cleanups in TINY_RCU priority boosting. rcu: Switch to this_cpu() primitives rcu: Use WARN_ON_ONCE for DEBUG_OBJECTS_RCU_HEAD warnings rcu: mark rcutorture boosting callback as being on-stack rcu: add DEBUG_OBJECTS_RCU_HEAD check for alignment rcu: Enable DEBUG_OBJECTS_RCU_HEAD from !PREEMPT rcu: Add forward-progress diagnostic for per-CPU kthreads rcu: add grace-period age and more kthread state to tracing rcu: fix tracing bug thinko on boost-balk attribution rcu: update tracing documentation for new rcutorture and rcuboost ... Pulling in rcu_virt_note_context_switch(). Signed-off-by:
-
Takuya Yoshikawa authored
We introduce em_jmp_far(). We also call this from em_grp45() to stop treating modrm_reg == 5 case separately in the group 5 emulation. Signed-off-by:
-
Takuya Yoshikawa authored
The prototypes are changed appropriately. We also replaces "goto grp45;" with simple em_grp45() call. Signed-off-by:
-
Takuya Yoshikawa authored
The opt of emulate_grp1a() is also removed. Signed-off-by:
-
Takuya Yoshikawa authored
Remove inline at this chance. Signed-off-by:
-
Takuya Yoshikawa authored
Signed-off-by:
-
Takuya Yoshikawa authored
In addition, one comma at the end of a statement is replaced with a semicolon. Signed-off-by:
-
Takuya Yoshikawa authored
This way, we can avoid checking the user space address many times when we read the guest memory. Although we can do the same for write if we check which slots are writable, we do not care write now: reading the guest memory happens more often than writing. [avi: change VERIFY_READ to VERIFY_WRITE] Signed-off-by:
-
Takuya Yoshikawa authored
When we optimized walk_addr_generic() by not using the generic guest memory reader, we replaced copy_from_user() with get_user(): commit e30d2a170506830d5eef5e9d7990c5aedf1b0a51 KVM: MMU: Optimize guest page table walk commit 15e2ac9a43d4d7d08088e404fddf2533a8e7d52e KVM: MMU: Fix 64-bit paging breakage on x86_32 But as Andi pointed out later, copy_from_user() does the same as get_user() as long as we give a constant size to it. So we use copy_from_user() to clean up the code. The only, noticeable, regression introduced by this is 64-bit gpte reading on x86_32 hosts needed for PAE guests. But this can be mitigated by implementing 8-byte get_user() for x86_32, if needed. Signed-off-by:
-
Scott Wood authored
Signed-off-by:
Scott Wood <scottwood@freescale.com> Signed-off-by:
Alexander Graf <agraf@suse.de>
-
Scott Wood authored
Linux doesn't use USPRG0 (now renamed VRSAVE in the architecture, even when Altivec isn't involved), but a guest might. Signed-off-by:
-
Stuart Yoder authored
Convert to microseconds when displaying (with fix from Bharat Bhushan <Bharat.Bhushan@freescale.com>). This reduces rounding error with large quantities of short exits. Signed-off-by:
Stuart Yoder <stuart.yoder@freescale.com> Signed-off-by:
-
Scott Wood authored
The exit type setting for mfspr/mtspr is moved from 44x to toplevel SPR emulation. This enables it on e500, and makes sure that all SPRs are covered. Exit accounting for tlbwe and tlbsx is added to e500. Signed-off-by:
-
Scott Wood authored
Return the actual host SVR for now, as we already do for PVR. Eventually we may support Qemu overriding PVR/SVR if the situation is appropriate, once we implement KVM_SET_SREGS on e500. Signed-off-by:
-
Avi Kivity authored
Since the emulator now checks segment limits and access rights, it generates a lot more accesses to the vmcs segment fields. Undo some of the performance hit by cacheing those fields in a read-only cache (the entire cache is invalidated on any write, or on guest exit). Signed-off-by: -
Avi Kivity authored
Instead of separate accessors for the segment selector and cached descriptor, use one accessor for both. This simplifies the code somewhat. Signed-off-by: -
Avi Kivity authored
Avoids a VMREAD. Signed-off-by: -
Joe Perches authored
dump_vmcb isn't used outside this module, make it static. Shrink text and object by ~1% by standardizing formats. $ size arch/x86/kvm/svm.o* text data bss dec hex filename 52910 580 10072 63562 f84a arch/x86/kvm/svm.o.new 53563 580 10072 64215 fad7 arch/x86/kvm/svm.o.old Signed-off-by:
Joe Perches <joe@perches.com> Acked-by:
Joerg Roedel <joerg.roedel@amd.com> Signed-off-by:
-
Takuya Yoshikawa authored
Fix regression introduced by commit e30d2a170506830d5eef5e9d7990c5aedf1b0a51 KVM: MMU: Optimize guest page table walk On x86_32, get_user() does not support 64-bit values and we fail to build KVM at the point of 64-bit paging. This patch fixes this by using get_user() twice for that condition. Signed-off-by:
Jan Kiszka <jan.kiszka@web.de> Signed-off-by:
-
Jeff Mahoney authored
This patch fixes some sparse warning about "dubious one-bit signed bitfield." Signed-off-by:
Jeff Mahoney <jeffm@suse.com> Originally-by:
Jan Blunck <jblunck@suse.de> Signed-off-by:
Xiantao Zhang <xiantao.zhang@intel.com> Signed-off-by:
-
BrillyWu@viatech.com.cn authored
The CPUIDs for Centaur are added, and then the features of PadLock hardware engine on VIA CPU, such as "ace", "ace_en" and so on, can be passed into the kvm guest. Signed-off-by:
Brilly Wu <brillywu@viatech.com.cn> Signed-off-by:
Kary Jin <karyjin@viatech.com.cn> Signed-off-by:
-
Gleb Natapov authored
Signed-off-by:
-
Gleb Natapov authored
mmio_index should be taken into account when copying data from userspace. Signed-off-by:
-
Gleb Natapov authored
Signed-off-by:
-
Gleb Natapov authored
Remove unused variable mmio_fault_cr2. Signed-off-by:
-
Avi Kivity authored
Move all groups into a single field and handle them in a single place. This saves bits when we add more group types (3 bits -> 7 groups types). Signed-off-by: -
Avi Kivity authored
walk_addr_generic() is a hot path and is also hard for the cpu to predict - some of the parameters (fetch_fault in particular) vary wildly from invocation to invocation. Add unlikely() annotations where appropriate; all walk failures are considered unlikely, as are cases where we have to mark the accessed or dirty bit, as they are slow paths both in kvm and on real processors. Signed-off-by: -
Takuya Yoshikawa authored
For this, em_pushf/popf() are introduced. Signed-off-by:
-
Takuya Yoshikawa authored
For this, emulate_pusha/popa() are converted to em_pusha/popa(). Signed-off-by:
-
Takuya Yoshikawa authored
In addition, the RET emulation is changed to call em_pop() to remove the pop_instruction label. Signed-off-by:
-
Takuya Yoshikawa authored
The following instructions are changed to use opcode::execute. Group 1 (80-83) ADD (00-05), OR (08-0D), ADC (10-15), SBB (18-1D), AND (20-25), SUB (28-2D), XOR (30-35), CMP (38-3D) CMPS (A6-A7), SCAS (AE-AF) The last two do the same as CMP in the emulator, so em_cmp() is used. Signed-off-by:
-
Takuya Yoshikawa authored
This patch optimizes the guest page table walk by using get_user() instead of copy_from_user(). With this patch applied, paging64_walk_addr_generic() has become about 0.5us to 1.0us faster on my Phenom II machine with NPT on. Signed-off-by:
-
Avi Kivity authored
By reserving 0 as an invalid x86_intercept_stage, we no longer need to store a valid flag in x86_intercept_map. Signed-off-by: -
Avi Kivity authored
Signed-off-by: -
Avi Kivity authored
While it isn't defined, no need to force a #UD. If it becomes defined in the future this can cause wierd problems for the guest. Signed-off-by: -
Avi Kivity authored
Signed-off-by:
-
