- 05 May, 2014 1 commit
-
-
Christian Borntraeger authored
When starting lots of dataplane devices the bootup takes very long on Christian's s390 with irqfd patches. With larger setups he is even able to trigger some timeouts in some components. Turns out that the KVM_SET_GSI_ROUTING ioctl takes very long (strace claims up to 0.1 sec) when having multiple CPUs. This is caused by the synchronize_rcu and the HZ=100 of s390. By changing the code to use a private srcu we can speed things up. This patch reduces the boot time till mounting root from 8 to 2 seconds on my s390 guest with 100 disks. Uses of hlist_for_each_entry_rcu, hlist_add_head_rcu, hlist_del_init_rcu are fine because they do not have lockdep checks (hlist_for_each_entry_rcu uses rcu_dereference_raw rather than rcu_dereference, and write-sides do not do rcu lockdep at all). Note that we're hardly relying on the "sleepable" part of srcu. We just want SRCU's faster detection of grace periods. Testing was done by Andrew Theurer using netperf tests STREAM, MAERTS and RR. The difference between results "before" and "after" the patch has mean -0.2% and standard deviation 0.6%. Using a paired t-test on the data points says that there is a 2.5% probability that the patch is the cause of the performance difference (rather than a random fluctuation). (Restricting the t-test to RR, which is the most likely to be affected, changes the numbers to respectively -0.3% mean, 0.7% stdev, and 8% probability that the numbers actually say something about the patch. The probability increases mostly because there are fewer data points). Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Michael S. Tsirkin <mst@redhat.com> Tested-by: Christian Borntraeger <borntraeger@de.ibm.com> # s390 Reviewed-by:
Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com>
-
- 30 Apr, 2014 1 commit
-
-
Paolo Bonzini authored
Merge tag 'kvm-s390-20140429' of git://git.kernel.org/pub/scm/linux/kernel/git/kvms390/linux into kvm-next 1. Guest handling fixes The handling of MVPG, PFMF and Test Block is fixed to better follow the architecture. None of these fixes is critical for any current Linux guests, but let's play safe. 2. Optimization for single CPU guests We can enable the IBS facility if only one VCPU is running (!STOPPED state). We also enable this optimization for guest > 1 VCPU as soon as all but one VCPU is in stopped state. Thus will help guests that have tools like cpuplugd (from s390-utils) that do dynamic offline/ online of CPUs. 3. NOTES There is one non-s390 change in include/linux/kvm_host.h that introduces 2 defines for VCPU requests: define KVM_REQ_ENABLE_IBS 23 define KVM_REQ_DISABLE_IBS 24
-
- 29 Apr, 2014 7 commits
-
-
Marcelo Tosatti authored
Invariant TSC is a property of TSC, no additional support code necessary. Signed-off-by:
Marcelo Tosatti <mtosatti@redhat.com> Signed-off-by:
-
David Hildenbrand authored
This patch enables the IBS facility when a single VCPU is running. The facility is dynamically turned on/off as soon as other VCPUs enter/leave the stopped state. When this facility is operating, some instructions can be executed faster for single-cpu guests. Signed-off-by:
David Hildenbrand <dahi@linux.vnet.ibm.com> Reviewed-by:
Dominik Dingel <dingel@linux.vnet.ibm.com> Reviewed-by:
Cornelia Huck <cornelia.huck@de.ibm.com> Signed-off-by:
-
David Hildenbrand authored
This patch introduces two new functions to set/clear the CPUSTAT_STOPPED bit and makes use of it at all applicable places. These functions prepare the additional execution of code when starting/stopping a vcpu. The CPUSTAT_STOPPED bit should not be touched outside of these functions. Signed-off-by:
-
Thomas Huth authored
TEST BLOCK is also subject to the low-address protection, so we need to check the destination address in our handler. Signed-off-by:
Thomas Huth <thuth@linux.vnet.ibm.com> Reviewed-by:
-
Thomas Huth authored
Add a check for low-address protection to the PFMF handler and convert real-addresses to absolute if necessary, as it is defined in the Principles of Operations specification. Signed-off-by:
-
Thomas Huth authored
The s390 architecture has a special protection mechanism that can be used to prevent write access to the vital data in the low-core memory area. This patch adds a new helper function that can be used to check for such write accesses and in case of protection, it also sets up the exception data accordingly. Signed-off-by:
-
Thomas Huth authored
When the guest executes the MVPG instruction with DAT disabled, and the source or destination page is not mapped in the host, the so-called partial execution interception occurs. We need to handle this event by setting up a mapping for the corresponding user pages. Signed-off-by:
-
- 28 Apr, 2014 2 commits
-
-
Oleg Nesterov authored
async_pf_execute() passes tsk == current to gup(), this is doesn't hurt but unnecessary and misleading. "tsk" is only used to account the number of faults and current is the random workqueue thread. Signed-off-by:
Oleg Nesterov <oleg@redhat.com> Suggested-by:
Andrea Arcangeli <aarcange@redhat.com> Signed-off-by:
-
Oleg Nesterov authored
async_pf_execute() has no reasons to adopt apf->mm, gup(current, mm) should work just fine even if current has another or NULL ->mm. Recently kvm_async_page_present_sync() was added insedie the "use_mm" section, but it seems that it doesn't need current->mm too. Signed-off-by:
-
- 23 Apr, 2014 9 commits
-
-
Xiao Guangrong authored
Now we can flush all the TLBs out of the mmu lock without TLB corruption when write-proect the sptes, it is because: - we have marked large sptes readonly instead of dropping them that means we just change the spte from writable to readonly so that we only need to care the case of changing spte from present to present (changing the spte from present to nonpresent will flush all the TLBs immediately), in other words, the only case we need to care is mmu_spte_update() - in mmu_spte_update(), we haved checked SPTE_HOST_WRITEABLE | PTE_MMU_WRITEABLE instead of PT_WRITABLE_MASK, that means it does not depend on PT_WRITABLE_MASK anymore Acked-by:
Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com> Signed-off-by:
-
Xiao Guangrong authored
Relax the tlb flush condition since we will write-protect the spte out of mmu lock. Note lockless write-protection only marks the writable spte to readonly and the spte can be writable only if both SPTE_HOST_WRITEABLE and SPTE_MMU_WRITEABLE are set (that are tested by spte_is_locklessly_modifiable) This patch is used to avoid this kind of race: VCPU 0 VCPU 1 lockless wirte protection: set spte.w = 0 lock mmu-lock write protection the spte to sync shadow page, see spte.w = 0, then without flush tlb unlock mmu-lock !!! At this point, the shadow page can still be writable due to the corrupt tlb entry Flush all TLB Reviewed-by: -
Xiao Guangrong authored
Currently, kvm zaps the large spte if write-protected is needed, the later read can fault on that spte. Actually, we can make the large spte readonly instead of making them un-present, the page fault caused by read access can be avoided The idea is from Avi: | As I mentioned before, write-protecting a large spte is a good idea, | since it moves some work from protect-time to fault-time, so it reduces | jitter. This removes the need for the return value. This version has fixed the issue reported in 6b73a960, the reason of that issue is that fast_page_fault() directly sets the readonly large spte to writable but only dirty the first page into the dirty-bitmap that means other pages are missed. Fixed it by only the normal sptes (on the PT_PAGE_TABLE_LEVEL level) can be fast fixed Reviewed-by:
-
Xiao Guangrong authored
Using sp->role.level instead of @level since @level is not got from the page table hierarchy There is no issue in current code since the fast page fault currently only fixes the fault caused by dirty-log that is always on the last level (level = 1) This patch makes the code more readable and avoids potential issue in the further development Reviewed-by:
-
Xiao Guangrong authored
This reverts commit 5befdc38. Since we will allow flush tlb out of mmu-lock in the later patch Signed-off-by:
-
Nadav Amit authored
If EFER.LMA is off, cs.l does not determine execution mode. Currently, the emulation engine assumes differently. Signed-off-by:
Nadav Amit <namit@cs.technion.ac.il> Signed-off-by:
-
Nadav Amit authored
The IN instruction is not be affected by REP-prefix as INS is. Therefore, the emulation should ignore the REP prefix as well. The current emulator implementation tries to perform writeback when IN instruction with REP-prefix is emulated. This causes it to perform wrong memory write or spurious #GP exception to be injected to the guest. Signed-off-by:
-
Nadav Amit authored
According to Intel specifications, PAE and non-PAE does not have any reserved bits. In long-mode, regardless to PCIDE, only the high bits (above the physical address) are reserved. Signed-off-by:
-
Nadav Amit authored
If a guest enables a performance counter but does not enable PMI, the hypervisor currently does not reprogram the performance counter once it overflows. As a result the host performance counter is kept with the original sampling period which was configured according to the value of the guest's counter when the counter was enabled. Such behaviour can cause very bad consequences. The most distrubing one can cause the guest not to make any progress at all, and keep exiting due to host PMI before any guest instructions is exeucted. This situation occurs when the performance counter holds a very high value when the guest enables the performance counter. As a result the host's sampling period is configured to be very short. The host then never reconfigures the sampling period and get stuck at entry->PMI->exit loop. We encountered such a scenario in our experiments. The solution is to reprogram the counter even if the guest does not use PMI. Signed-off-by:
-
- 22 Apr, 2014 20 commits
-
-
Bandan Das authored
Some Type 1 hypervisors such as XEN won't enable VMX without it present Signed-off-by:
Bandan Das <bsd@redhat.com> Signed-off-by:
-
Bandan Das authored
This feature emulates the "Acknowledge interrupt on exit" behavior. We can safely emulate it for L1 to run L2 even if L0 itself has it disabled (to run L1). Signed-off-by:
-
Bandan Das authored
For single context invalidation, we fall through to global invalidation in handle_invept() except for one case - when the operand supplied by L1 is different from what we have in vmcs12. However, typically hypervisors will only call invept for the currently loaded eptp, so the condition will never be true. Signed-off-by:
-
Huw Davies authored
When entering an exception after an ICEBP, the saved instruction pointer should point to after the instruction. This fixes the bug here: https://bugs.launchpad.net/qemu/+bug/1119686Signed-off-by:
Huw Davies <huw@codeweavers.com> Reviewed-by:
Jan Kiszka <jan.kiszka@siemens.com> Signed-off-by:
-
Marcelo Tosatti authored
Merge tag 'kvm-s390-20140422' of git://git.kernel.org/pub/scm/linux/kernel/git/kvms390/linux into queue Lazy storage key handling ------------------------- Linux does not use the ACC and F bits of the storage key. Newer Linux versions also do not use the storage keys for dirty and reference tracking. We can optimize the guest handling for those guests for faults as well as page-in and page-out by simply not caring about the guest visible storage key. We trap guest storage key instruction to enable those keys only on demand. Migration bitmap Until now s390 never provided a proper dirty bitmap. Let's provide a proper migration bitmap for s390. We also change the user dirty tracking to a fault based mechanism. This makes the host completely independent from the storage keys. Long term this will allow us to back guest memory with large pages. per-VM device attributes ------------------------ To avoid the introduction of new ioctls, let's provide the attribute semanantic also on the VM-"device". Userspace controlled CMMA ------------------------- The CMMA assist is changed from "always on" to "on if requested" via per-VM device attributes. In addition a callback to reset all usage states is provided. Proper guest DAT handling for intercepts ---------------------------------------- While instructions handled by SIE take care of all addressing aspects, KVM/s390 currently does not care about guest address translation of intercepts. This worked out fine, because - the s390 Linux kernel has a 1:1 mapping between kernel virtual<->real for all pages up to memory size - intercepts happen only for a small amount of cases - all of these intercepts happen to be in the kernel text for current distros Of course we need to be better for other intercepts, kernel modules etc. We provide the infrastructure and rework all in-kernel intercepts to work on logical addresses (paging etc) instead of real ones. The code has been running internally for several months now, so it is time for going public. GDB support ----------- We provide breakpoints, single stepping and watchpoints. Fixes/Cleanups -------------- - Improve program check delivery - Factor out the handling of transactional memory on program checks - Use the existing define __LC_PGM_TDB - Several cleanups in the lowcore structure - Documentation NOTES ----- - All patches touching base s390 are either ACKed or written by the s390 maintainers - One base KVM patch "KVM: add kvm_is_error_gpa() helper" - One patch introduces the notion of VM device attributes Signed-off-by:
-
Michael Mueller authored
Factor out the new function handle_itdb(), which copies the ITDB into guest lowcore to fully handle a TX abort. Signed-off-by:
Michael Mueller <mimu@linux.vnet.ibm.com> Acked-by:
-
Michael Mueller authored
The generically assembled low core labels already contain the address for the TDB. Signed-off-by:
-
Christian Borntraeger authored
On hard exits (abort, sigkill) we have have some kvm_s390_interrupt_info structures hanging around. Delete those on exit to avoid memory leaks. Signed-off-by:
-
David Hildenbrand authored
When a guest is single-stepped, we want to disable timer interrupts. Otherwise, the guest will continuously execute the external interrupt handler and make debugging of code where timer interrupts are enabled almost impossible. The delivery of timer interrupts can be enforced in such sections by setting a breakpoint and continuing execution. In order to disable timer interrupts, they are disabled in the control register of the guest just before SIE entry and are suppressed in the interrupt check/delivery methods. Signed-off-by:
-
David Hildenbrand authored
This patch moves the checks for enabled timer (clock-comparator) interrupts and pending timer interrupts into own functions, making the code better readable and easier to maintain. The method kvm_cpu_has_pending_timer is filled with life. Signed-off-by:
-
David Hildenbrand authored
Added documentation for diag 501, stating that no subfunctions are provided and no parameters are used. Signed-off-by:
-
David Hildenbrand authored
This patch adds support to debug the guest using the PER facility on s390. Single-stepping, hardware breakpoints and hardware watchpoints are supported. In order to use the PER facility of the guest without it noticing it, the control registers of the guest have to be patched and access to them has to be intercepted(stctl, stctg, lctl, lctlg). All PER program interrupts have to be intercepted and only the relevant PER interrupts for the guest have to be given back. Special care has to be taken about repeated exits on the same hardware breakpoint. The intervention of the host in the guests PER configuration is not fully transparent. PER instruction nullification can not be used by the guest and too many storage alteration events may be reported to the guest (if it is activated for special address ranges only) when the host concurrently debugging it. Signed-off-by:
-
David Hildenbrand authored
This patch adds the structs to the kernel headers needed to pass information from/to userspace in order to debug a guest on s390 with hardware support. Signed-off-by:
-
David Hildenbrand authored
Introduce the methods to emulate the stctl and stctg instruction. Added tracing code. Signed-off-by:
-
David Hildenbrand authored
When a program interrupt was to be delivered until now, no program interrupt parameters were stored in the low-core of the target vcpu. This patch enables the delivery of those program interrupt parameters, takes care of concurrent PER events which can be injected in addition to any program interrupt and uses the correct instruction length code (depending on the interception code) for the injection of program interrupts. Signed-off-by:
-
David Hildenbrand authored
Whenever a program interrupt is intercepted, some parameters are stored in the sie control block. These parameters have to be extracted in order to be reinjected correctly. This patch also takes care of intercepted PER events which can occurr in addition to any program interrupt. Signed-off-by:
-
Jens Freimann authored
This patch adds fields which are currently missing but needed for the correct injection of interrupts. This is based on a patch by David Hildenbrand Signed-off-by:
Jens Freimann <jfrei@linux.vnet.ibm.com> Acked-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by:
-
Jens Freimann authored
per_perc_atmid is currently a two-byte field that combines two fields, the PER code and the PER Addressing-and-Translation-Mode Identification (ATMID) Let's make them accessible indepently and also rename per_cause to per_code. Signed-off-by:
-
Jens Freimann authored
According to the Principles of Operation, at offset 0xA3 in the lowcore we have the "Architectural-Mode identification", not an "access identification". Signed-off-by:
-
Heiko Carstens authored
Add kerneldoc comment to kvm_s390_real_to_abs() and change the code so it matches the coding style of the rest of gaccess.h. Signed-off-by:
-
