kernel/linux.git/arch/x86/hyperv, branch v5.15.208

x86/hyperv: Remove unregister syscore call from Hyper-V cleanup

2022-12-31T12:14:39+00:00

[ Upstream commit 32c97d980e2eef25465d453f2956a9ca68926a3c ] Hyper-V cleanup code comes under panic path where preemption and irq is already disabled. So calling of unregister_syscore_ops might schedule out the thread even for the case where mutex lock is free. hyperv_cleanup unregister_syscore_ops mutex_lock(&syscore_ops_lock) might_sleep Here might_sleep might schedule out this thread, where voluntary preemption config is on and this thread will never comes back. And also this was added earlier to maintain the symmetry which is not required as this can comes during crash shutdown path only. To prevent the same, removing unregister_syscore_ops function call. Signed-off-by: Gaurav Kohli Reviewed-by: Michael Kelley Link: https://lore.kernel.org/r/1669443291-2575-1-git-send-email-gauravkohli@linux.microsoft.com Signed-off-by: Wei Liu Signed-off-by: Sasha Levin

x86/hyperv: Restore VP assist page after cpu offlining/onlining

2022-12-02T16:41:03+00:00

[ Upstream commit ee6815416380bc069b7dcbdff0682d4c53617527 ] Commit e5d9b714fe40 ("x86/hyperv: fix root partition faults when writing to VP assist page MSR") moved 'wrmsrl(HV_X64_MSR_VP_ASSIST_PAGE)' under 'if (*hvp)' condition. This works for root partition as hv_cpu_die() does memunmap() and sets 'hv_vp_assist_page[cpu]' to NULL but breaks non-root partitions as hv_cpu_die() doesn't free 'hv_vp_assist_page[cpu]' for them. This causes VP assist page to remain unset after CPU offline/online cycle: $ rdmsr -p 24 0x40000073 10212f001 $ echo 0 > /sys/devices/system/cpu/cpu24/online $ echo 1 > /sys/devices/system/cpu/cpu24/online $ rdmsr -p 24 0x40000073 0 Fix the issue by always writing to HV_X64_MSR_VP_ASSIST_PAGE in hv_cpu_init(). Note, checking 'if (!*hvp)', for root partition is pointless as hv_cpu_die() always sets 'hv_vp_assist_page[cpu]' to NULL (and it's also NULL initially). Note: the fact that 'hv_vp_assist_page[cpu]' is reset to NULL may present a (potential) issue for KVM. While Hyper-V uses CPUHP_AP_ONLINE_DYN stage in CPU hotplug, KVM uses CPUHP_AP_KVM_STARTING which comes earlier in CPU teardown sequence. It is theoretically possible that Enlightened VMCS is still in use. It is unclear if the issue is real and if using KVM with Hyper-V root partition is even possible. While on it, drop the unneeded smp_processor_id() call from hv_cpu_init(). Fixes: e5d9b714fe40 ("x86/hyperv: fix root partition faults when writing to VP assist page MSR") Signed-off-by: Vitaly Kuznetsov Reviewed-by: Michael Kelley Link: https://lore.kernel.org/r/20221103190601.399343-1-vkuznets@redhat.com Signed-off-by: Wei Liu Signed-off-by: Sasha Levin

x86/hyperv: Properly deal with empty cpumasks in hyperv_flush_tlb_multi()

2022-03-08T18:12:36+00:00

[ Upstream commit 51500b71d500f251037ed339047a4d9e7d7e295b ] KASAN detected the following issue: BUG: KASAN: slab-out-of-bounds in hyperv_flush_tlb_multi+0xf88/0x1060 Read of size 4 at addr ffff8880011ccbc0 by task kcompactd0/33 CPU: 1 PID: 33 Comm: kcompactd0 Not tainted 5.14.0-39.el9.x86_64+debug #1 Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS Hyper-V UEFI Release v4.0 12/17/2019 Call Trace: dump_stack_lvl+0x57/0x7d print_address_description.constprop.0+0x1f/0x140 ? hyperv_flush_tlb_multi+0xf88/0x1060 __kasan_report.cold+0x7f/0x11e ? hyperv_flush_tlb_multi+0xf88/0x1060 kasan_report+0x38/0x50 hyperv_flush_tlb_multi+0xf88/0x1060 flush_tlb_mm_range+0x1b1/0x200 ptep_clear_flush+0x10e/0x150 ... Allocated by task 0: kasan_save_stack+0x1b/0x40 __kasan_kmalloc+0x7c/0x90 hv_common_init+0xae/0x115 hyperv_init+0x97/0x501 apic_intr_mode_init+0xb3/0x1e0 x86_late_time_init+0x92/0xa2 start_kernel+0x338/0x3eb secondary_startup_64_no_verify+0xc2/0xcb The buggy address belongs to the object at ffff8880011cc800 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 960 bytes inside of 1024-byte region [ffff8880011cc800, ffff8880011ccc00) 'hyperv_flush_tlb_multi+0xf88/0x1060' points to hv_cpu_number_to_vp_number() and '960 bytes' means we're trying to get VP_INDEX for CPU#240. 'nr_cpus' here is exactly 240 so we're trying to access past hv_vp_index's last element. This can (and will) happen when 'cpus' mask is empty and cpumask_last() will return '>=nr_cpus'. Commit ad0a6bad4475 ("x86/hyperv: check cpu mask after interrupt has been disabled") tried to deal with empty cpumask situation but apparently didn't fully fix the issue. 'cpus' cpumask which is passed to hyperv_flush_tlb_multi() is 'mm_cpumask(mm)' (which is '&mm->cpu_bitmap'). This mask changes every time the particular mm is scheduled/unscheduled on some CPU (see switch_mm_irqs_off()), disabling IRQs on the CPU which is performing remote TLB flush has zero influence on whether the particular process can get scheduled/unscheduled on _other_ CPUs so e.g. in the case where the mm was scheduled on one other CPU and got unscheduled during hyperv_flush_tlb_multi()'s execution will lead to cpumask becoming empty. It doesn't seem that there's a good way to protect 'mm_cpumask(mm)' from changing during hyperv_flush_tlb_multi()'s execution. It would be possible to copy it in the very beginning of the function but this is a waste. It seems we can deal with changing cpumask just fine. When 'cpus' cpumask changes during hyperv_flush_tlb_multi()'s execution, there are two possible issues: - 'Under-flushing': we will not flush TLB on a CPU which got added to the mask while hyperv_flush_tlb_multi() was already running. This is not a problem as this is equal to mm getting scheduled on that CPU right after TLB flush. - 'Over-flushing': we may flush TLB on a CPU which is already cleared from the mask. First, extra TLB flush preserves correctness. Second, Hyper-V's TLB flush hypercall takes 'mm->pgd' argument so Hyper-V may avoid the flush if CR3 doesn't match. Fix the immediate issue with cpumask_last()/hv_cpu_number_to_vp_number() and remove the pointless cpumask_empty() check from the beginning of the function as it really doesn't protect anything. Also, avoid the hypercall altogether when 'flush->processor_mask' ends up being empty. Fixes: ad0a6bad4475 ("x86/hyperv: check cpu mask after interrupt has been disabled") Signed-off-by: Vitaly Kuznetsov Reviewed-by: Michael Kelley Link: https://lore.kernel.org/r/20220106094611.1404218-1-vkuznets@redhat.com Signed-off-by: Wei Liu Signed-off-by: Sasha Levin

x86/hyperv: Move required MSRs check to initial platform probing

2021-12-08T08:04:38+00:00

[ Upstream commit f3e613e72f66226b3bea1046c1b864f67a3000a4 ] Explicitly check for MSR_HYPERCALL and MSR_VP_INDEX support when probing for running as a Hyper-V guest instead of waiting until hyperv_init() to detect the bogus configuration. Add messages to give the admin a heads up that they are likely running on a broken virtual machine setup. At best, silently disabling Hyper-V is confusing and difficult to debug, e.g. the kernel _says_ it's using all these fancy Hyper-V features, but always falls back to the native versions. At worst, the half baked setup will crash/hang the kernel. Reviewed-by: Vitaly Kuznetsov Signed-off-by: Sean Christopherson Link: https://lore.kernel.org/r/20211104182239.1302956-3-seanjc@google.com Signed-off-by: Wei Liu Signed-off-by: Sasha Levin

x86/hyperv: Fix NULL deref in set_hv_tscchange_cb() if Hyper-V setup fails

2021-11-25T08:48:43+00:00

commit daf972118c517b91f74ff1731417feb4270625a4 upstream. Check for a valid hv_vp_index array prior to derefencing hv_vp_index when setting Hyper-V's TSC change callback. If Hyper-V setup failed in hyperv_init(), the kernel will still report that it's running under Hyper-V, but will have silently disabled nearly all functionality. BUG: kernel NULL pointer dereference, address: 0000000000000010 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 0 P4D 0 Oops: 0000 [#1] SMP CPU: 4 PID: 1 Comm: swapper/0 Not tainted 5.15.0-rc2+ #75 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 0.0.0 02/06/2015 RIP: 0010:set_hv_tscchange_cb+0x15/0xa0 Code: <8b> 04 82 8b 15 12 17 85 01 48 c1 e0 20 48 0d ee 00 01 00 f6 c6 08 ... Call Trace: kvm_arch_init+0x17c/0x280 kvm_init+0x31/0x330 vmx_init+0xba/0x13a do_one_initcall+0x41/0x1c0 kernel_init_freeable+0x1f2/0x23b kernel_init+0x16/0x120 ret_from_fork+0x22/0x30 Fixes: 93286261de1b ("x86/hyperv: Reenlightenment notifications support") Cc: stable@vger.kernel.org Cc: Vitaly Kuznetsov Signed-off-by: Sean Christopherson Reviewed-by: Vitaly Kuznetsov Link: https://lore.kernel.org/r/20211104182239.1302956-2-seanjc@google.com Signed-off-by: Wei Liu Signed-off-by: Greg Kroah-Hartman

x86/hyperv: Protect set_hv_tscchange_cb() against getting preempted

2021-11-18T18:16:20+00:00

[ Upstream commit 285f68afa8b20f752b0b7194d54980b5e0e27b75 ] The following issue is observed with CONFIG_DEBUG_PREEMPT when KVM loads: KVM: vmx: using Hyper-V Enlightened VMCS BUG: using smp_processor_id() in preemptible [00000000] code: systemd-udevd/488 caller is set_hv_tscchange_cb+0x16/0x80 CPU: 1 PID: 488 Comm: systemd-udevd Not tainted 5.15.0-rc5+ #396 Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS Hyper-V UEFI Release v4.0 12/17/2019 Call Trace: dump_stack_lvl+0x6a/0x9a check_preemption_disabled+0xde/0xe0 ? kvm_gen_update_masterclock+0xd0/0xd0 [kvm] set_hv_tscchange_cb+0x16/0x80 kvm_arch_init+0x23f/0x290 [kvm] kvm_init+0x30/0x310 [kvm] vmx_init+0xaf/0x134 [kvm_intel] ... set_hv_tscchange_cb() can get preempted in between acquiring smp_processor_id() and writing to HV_X64_MSR_REENLIGHTENMENT_CONTROL. This is not an issue by itself: HV_X64_MSR_REENLIGHTENMENT_CONTROL is a partition-wide MSR and it doesn't matter which particular CPU will be used to receive reenlightenment notifications. The only real problem can (in theory) be observed if the CPU whose id was acquired with smp_processor_id() goes offline before we manage to write to the MSR, the logic in hv_cpu_die() won't be able to reassign it correctly. Reported-by: Michael Kelley Signed-off-by: Vitaly Kuznetsov Link: https://lore.kernel.org/r/20211012155005.1613352-1-vkuznets@redhat.com Signed-off-by: Wei Liu Signed-off-by: Sasha Levin

Merge tag 'hyperv-fixes-signed-20211007' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux

2021-10-07T16:44:48+00:00

Pull hyperv fixes from Wei Liu: - Replace uuid.h with types.h in a header (Andy Shevchenko) - Avoid sleeping in atomic context in PCI driver (Long Li) - Avoid sending IPI to self when it shouldn't (Vitaly Kuznetsov) * tag 'hyperv-fixes-signed-20211007' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux: x86/hyperv: Avoid erroneously sending IPI to 'self' hyper-v: Replace uuid.h with types.h PCI: hv: Fix sleep while in non-sleep context when removing child devices from the bus

x86/hyperv: Avoid erroneously sending IPI to 'self'

2021-10-06T15:56:45+00:00

__send_ipi_mask_ex() uses an optimization: when the target CPU mask is equal to 'cpu_present_mask' it uses 'HV_GENERIC_SET_ALL' format to avoid converting the specified cpumask to VP_SET. This case was overlooked when 'exclude_self' parameter was added. As the result, a spurious IPI to 'self' can be send. Reported-by: Thomas Gleixner Fixes: dfb5c1e12c28 ("x86/hyperv: remove on-stack cpumask from hv_send_ipi_mask_allbutself") Signed-off-by: Vitaly Kuznetsov Reviewed-by: Michael Kelley Link: https://lore.kernel.org/r/20211006125016.941616-1-vkuznets@redhat.com Signed-off-by: Wei Liu

Merge tag 'hyperv-fixes-signed-20210915' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux

2021-09-16T00:18:56+00:00

Pull hyperv fixes from Wei Liu: - Fix kernel crash caused by uio driver (Vitaly Kuznetsov) - Remove on-stack cpumask from HV APIC code (Wei Liu) * tag 'hyperv-fixes-signed-20210915' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux: x86/hyperv: remove on-stack cpumask from hv_send_ipi_mask_allbutself asm-generic/hyperv: provide cpumask_to_vpset_noself Drivers: hv: vmbus: Fix kernel crash upon unbinding a device from uio_hv_generic driver

x86/hyperv: remove on-stack cpumask from hv_send_ipi_mask_allbutself

2021-09-11T15:41:00+00:00

It is not a good practice to allocate a cpumask on stack, given it may consume up to 1 kilobytes of stack space if the kernel is configured to have 8192 cpus. The internal helper functions __send_ipi_mask{,_ex} need to loop over the provided mask anyway, so it is not too difficult to skip `self' there. We can thus do away with the on-stack cpumask in hv_send_ipi_mask_allbutself. Adjust call sites of __send_ipi_mask as needed. Reported-by: Linus Torvalds Suggested-by: Michael Kelley Suggested-by: Linus Torvalds Fixes: 68bb7bfb7985d ("X86/Hyper-V: Enable IPI enlightenments") Signed-off-by: Wei Liu Reviewed-by: Michael Kelley Link: https://lore.kernel.org/r/20210910185714.299411-3-wei.liu@kernel.org