linux-2.6-microblaze.git
2 years agoKVM: selftests: Make hyperv_clock selftest more stable
Vitaly Kuznetsov [Wed, 1 Jun 2022 14:43:22 +0000 (16:43 +0200)]
KVM: selftests: Make hyperv_clock selftest more stable

hyperv_clock doesn't always give a stable test result, especially with
AMD CPUs. The test compares Hyper-V MSR clocksource (acquired either
with rdmsr() from within the guest or KVM_GET_MSRS from the host)
against rdtsc(). To increase the accuracy, increase the measured delay
(done with nop loop) by two orders of magnitude and take the mean rdtsc()
value before and after rdmsr()/KVM_GET_MSRS.

Reported-by: Maxim Levitsky <mlevitsk@redhat.com>
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>
Tested-by: Maxim Levitsky <mlevitsk@redhat.com>
Message-Id: <20220601144322.1968742-1-vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86/MMU: Zap non-leaf SPTEs when disabling dirty logging
Ben Gardon [Wed, 25 May 2022 23:09:04 +0000 (23:09 +0000)]
KVM: x86/MMU: Zap non-leaf SPTEs when disabling dirty logging

Currently disabling dirty logging with the TDP MMU is extremely slow.
On a 96 vCPU / 96G VM backed with gigabyte pages, it takes ~200 seconds
to disable dirty logging with the TDP MMU, as opposed to ~4 seconds with
the shadow MMU.

When disabling dirty logging, zap non-leaf parent entries to allow
replacement with huge pages instead of recursing and zapping all of the
child, leaf entries. This reduces the number of TLB flushes required.
and reduces the disable dirty log time with the TDP MMU to ~3 seconds.

Opportunistically add a WARN() to catch GFNs that are mapped at a
higher level than their max level.

Signed-off-by: Ben Gardon <bgardon@google.com>
Message-Id: <20220525230904.1584480-1-bgardon@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agox86: drop bogus "cc" clobber from __try_cmpxchg_user_asm()
Jan Beulich [Tue, 7 Jun 2022 15:00:53 +0000 (17:00 +0200)]
x86: drop bogus "cc" clobber from __try_cmpxchg_user_asm()

As noted (and fixed) a couple of times in the past, "=@cc<cond>" outputs
and clobbering of "cc" don't work well together. The compiler appears to
mean to reject such, but doesn't - in its upstream form - quite manage
to yet for "cc". Furthermore two similar macros don't clobber "cc", and
clobbering "cc" is pointless in asm()-s for x86 anyway - the compiler
always assumes status flags to be clobbered there.

Fixes: 989b5db215a2 ("x86/uaccess: Implement macros for CMPXCHG on user addresses")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Message-Id: <485c0c0b-a3a7-0b7c-5264-7d00c01de032@suse.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86/mmu: Check every prev_roots in __kvm_mmu_free_obsolete_roots()
Shaoqin Huang [Tue, 7 Jun 2022 00:59:05 +0000 (18:59 -0600)]
KVM: x86/mmu: Check every prev_roots in __kvm_mmu_free_obsolete_roots()

When freeing obsolete previous roots, check prev_roots as intended, not
the current root.

Signed-off-by: Shaoqin Huang <shaoqin.huang@intel.com>
Fixes: 527d5cd7eece ("KVM: x86/mmu: Zap only obsolete roots if a root shadow page is zapped")
Message-Id: <20220607005905.2933378-1-shaoqin.huang@intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoentry/kvm: Exit to user mode when TIF_NOTIFY_SIGNAL is set
Seth Forshee [Wed, 4 May 2022 18:08:40 +0000 (13:08 -0500)]
entry/kvm: Exit to user mode when TIF_NOTIFY_SIGNAL is set

A livepatch transition may stall indefinitely when a kvm vCPU is heavily
loaded. To the host, the vCPU task is a user thread which is spending a
very long time in the ioctl(KVM_RUN) syscall. During livepatch
transition, set_notify_signal() will be called on such tasks to
interrupt the syscall so that the task can be transitioned. This
interrupts guest execution, but when xfer_to_guest_mode_work() sees that
TIF_NOTIFY_SIGNAL is set but not TIF_SIGPENDING it concludes that an
exit to user mode is unnecessary, and guest execution is resumed without
transitioning the task for the livepatch.

This handling of TIF_NOTIFY_SIGNAL is incorrect, as set_notify_signal()
is expected to break tasks out of interruptible kernel loops and cause
them to return to userspace. Change xfer_to_guest_mode_work() to handle
TIF_NOTIFY_SIGNAL the same as TIF_SIGPENDING, signaling to the vCPU run
loop that an exit to userpsace is needed. Any pending task_work will be
run when get_signal() is called from exit_to_user_mode_loop(), so there
is no longer any need to run task work from xfer_to_guest_mode_work().

Suggested-by: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Petr Mladek <pmladek@suse.com>
Signed-off-by: Seth Forshee <sforshee@digitalocean.com>
Message-Id: <20220504180840.2907296-1-sforshee@digitalocean.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: Don't null dereference ops->destroy
Alexey Kardashevskiy [Wed, 1 Jun 2022 01:43:28 +0000 (03:43 +0200)]
KVM: Don't null dereference ops->destroy

A KVM device cleanup happens in either of two callbacks:
1) destroy() which is called when the VM is being destroyed;
2) release() which is called when a device fd is closed.

Most KVM devices use 1) but Book3s's interrupt controller KVM devices
(XICS, XIVE, XIVE-native) use 2) as they need to close and reopen during
the machine execution. The error handling in kvm_ioctl_create_device()
assumes destroy() is always defined which leads to NULL dereference as
discovered by Syzkaller.

This adds a checks for destroy!=NULL and adds a missing release().

This is not changing kvm_destroy_devices() as devices with defined
release() should have been removed from the KVM devices list by then.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86: Fix the intel_pt PMI handling wrongly considered from guest
Yanfei Xu [Mon, 23 May 2022 14:08:21 +0000 (22:08 +0800)]
KVM: x86: Fix the intel_pt PMI handling wrongly considered from guest

When kernel handles the vm-exit caused by external interrupts and NMI,
it always sets kvm_intr_type to tell if it's dealing an IRQ or NMI. For
the PMI scenario, it could be IRQ or NMI.

However, intel_pt PMIs are only generated for HARDWARE perf events, and
HARDWARE events are always configured to generate NMIs.  Use
kvm_handling_nmi_from_guest() to precisely identify if the intel_pt PMI
came from the guest; this avoids false positives if an intel_pt PMI/NMI
arrives while the host is handling an unrelated IRQ VM-Exit.

Fixes: db215756ae59 ("KVM: x86: More precisely identify NMI from guest when handling PMI")
Signed-off-by: Yanfei Xu <yanfei.xu@intel.com>
Message-Id: <20220523140821.1345605-1-yanfei.xu@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: selftests: x86: Sync the new name of the test case to .gitignore
Like Xu [Wed, 18 May 2022 17:01:17 +0000 (01:01 +0800)]
KVM: selftests: x86: Sync the new name of the test case to .gitignore

Fixing side effect of the so-called opportunistic change in the commit.

Fixes: dc8a9febbab0 ("KVM: selftests: x86: Fix test failure on arch lbr capable platforms")
Signed-off-by: Like Xu <likexu@tencent.com>
Message-Id: <20220518170118.66263-2-likexu@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoDocumentation: kvm: reorder ARM-specific section about KVM_SYSTEM_EVENT_SUSPEND
Paolo Bonzini [Fri, 20 May 2022 11:22:46 +0000 (07:22 -0400)]
Documentation: kvm: reorder ARM-specific section about KVM_SYSTEM_EVENT_SUSPEND

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agox86, kvm: use correct GFP flags for preemption disabled
Paolo Bonzini [Tue, 24 May 2022 13:43:31 +0000 (09:43 -0400)]
x86, kvm: use correct GFP flags for preemption disabled

Commit ddd7ed842627 ("x86/kvm: Alloc dummy async #PF token outside of
raw spinlock") leads to the following Smatch static checker warning:

arch/x86/kernel/kvm.c:212 kvm_async_pf_task_wake()
warn: sleeping in atomic context

arch/x86/kernel/kvm.c
    202         raw_spin_lock(&b->lock);
    203         n = _find_apf_task(b, token);
    204         if (!n) {
    205                 /*
    206                  * Async #PF not yet handled, add a dummy entry for the token.
    207                  * Allocating the token must be down outside of the raw lock
    208                  * as the allocator is preemptible on PREEMPT_RT kernels.
    209                  */
    210                 if (!dummy) {
    211                         raw_spin_unlock(&b->lock);
--> 212                         dummy = kzalloc(sizeof(*dummy), GFP_KERNEL);
                                                                ^^^^^^^^^^
Smatch thinks the caller has preempt disabled.  The `smdb.py preempt
kvm_async_pf_task_wake` output call tree is:

sysvec_kvm_asyncpf_interrupt() <- disables preempt
-> __sysvec_kvm_asyncpf_interrupt()
   -> kvm_async_pf_task_wake()

The caller is this:

arch/x86/kernel/kvm.c
   290        DEFINE_IDTENTRY_SYSVEC(sysvec_kvm_asyncpf_interrupt)
   291        {
   292                struct pt_regs *old_regs = set_irq_regs(regs);
   293                u32 token;
   294
   295                ack_APIC_irq();
   296
   297                inc_irq_stat(irq_hv_callback_count);
   298
   299                if (__this_cpu_read(apf_reason.enabled)) {
   300                        token = __this_cpu_read(apf_reason.token);
   301                        kvm_async_pf_task_wake(token);
   302                        __this_cpu_write(apf_reason.token, 0);
   303                        wrmsrl(MSR_KVM_ASYNC_PF_ACK, 1);
   304                }
   305
   306                set_irq_regs(old_regs);
   307        }

The DEFINE_IDTENTRY_SYSVEC() is a wrapper that calls this function
from the call_on_irqstack_cond().  It's inside the call_on_irqstack_cond()
where preempt is disabled (unless it's already disabled).  The
irq_enter/exit_rcu() functions disable/enable preempt.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: LAPIC: Drop pending LAPIC timer injection when canceling the timer
Wanpeng Li [Fri, 20 May 2022 14:15:18 +0000 (10:15 -0400)]
KVM: LAPIC: Drop pending LAPIC timer injection when canceling the timer

The timer is disarmed when switching between TSC deadline and other modes;
however, the pending timer is still in-flight, so let's accurately remove
any traces of the previous mode.

Fixes: 4427593258 ("KVM: x86: thoroughly disarm LAPIC timer around TSC deadline switch")
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agox86/kvm: Alloc dummy async #PF token outside of raw spinlock
Sean Christopherson [Thu, 19 May 2022 14:57:11 +0000 (07:57 -0700)]
x86/kvm: Alloc dummy async #PF token outside of raw spinlock

Drop the raw spinlock in kvm_async_pf_task_wake() before allocating the
the dummy async #PF token, the allocator is preemptible on PREEMPT_RT
kernels and must not be called from truly atomic contexts.

Opportunistically document why it's ok to loop on allocation failure,
i.e. why the function won't get stuck in an infinite loop.

Reported-by: Yajun Deng <yajun.deng@linux.dev>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86: avoid calling x86 emulator without a decoded instruction
Sean Christopherson [Fri, 11 Mar 2022 03:27:41 +0000 (03:27 +0000)]
KVM: x86: avoid calling x86 emulator without a decoded instruction

Whenever x86_decode_emulated_instruction() detects a breakpoint, it
returns the value that kvm_vcpu_check_breakpoint() writes into its
pass-by-reference second argument.  Unfortunately this is completely
bogus because the expected outcome of x86_decode_emulated_instruction
is an EMULATION_* value.

Then, if kvm_vcpu_check_breakpoint() does "*r = 0" (corresponding to
a KVM_EXIT_DEBUG userspace exit), it is misunderstood as EMULATION_OK
and x86_emulate_instruction() is called without having decoded the
instruction.  This causes various havoc from running with a stale
emulation context.

The fix is to move the call to kvm_vcpu_check_breakpoint() where it was
before commit 4aa2691dcbd3 ("KVM: x86: Factor out x86 instruction
emulation with decoding") introduced x86_decode_emulated_instruction().
The other caller of the function does not need breakpoint checks,
because it is invoked as part of a vmexit and the processor has already
checked those before executing the instruction that #GP'd.

This fixes CVE-2022-1852.

Reported-by: Qiuhao Li <qiuhao@sysec.org>
Reported-by: Gaoning Pan <pgn@zju.edu.cn>
Reported-by: Yongkang Jia <kangel@zju.edu.cn>
Fixes: 4aa2691dcbd3 ("KVM: x86: Factor out x86 instruction emulation with decoding")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220311032801.3467418-2-seanjc@google.com>
[Rewrote commit message according to Qiuhao's report, since a patch
 already existed to fix the bug. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: SVM: Use kzalloc for sev ioctl interfaces to prevent kernel data leak
Ashish Kalra [Mon, 16 May 2022 15:43:10 +0000 (15:43 +0000)]
KVM: SVM: Use kzalloc for sev ioctl interfaces to prevent kernel data leak

For some sev ioctl interfaces, the length parameter that is passed maybe
less than or equal to SEV_FW_BLOB_MAX_SIZE, but larger than the data
that PSP firmware returns. In this case, kmalloc will allocate memory
that is the size of the input rather than the size of the data.
Since PSP firmware doesn't fully overwrite the allocated buffer, these
sev ioctl interface may return uninitialized kernel slab memory.

Reported-by: Andy Nguyen <theflow@google.com>
Suggested-by: David Rientjes <rientjes@google.com>
Suggested-by: Peter Gonda <pgonda@google.com>
Cc: kvm@vger.kernel.org
Cc: stable@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Fixes: eaf78265a4ab3 ("KVM: SVM: Move SEV code to separate file")
Fixes: 2c07ded06427d ("KVM: SVM: add support for SEV attestation command")
Fixes: 4cfdd47d6d95a ("KVM: SVM: Add KVM_SEV SEND_START command")
Fixes: d3d1af85e2c75 ("KVM: SVM: Add KVM_SEND_UPDATE_DATA command")
Fixes: eba04b20e4861 ("KVM: x86: Account a variety of miscellaneous allocations")
Signed-off-by: Ashish Kalra <ashish.kalra@amd.com>
Reviewed-by: Peter Gonda <pgonda@google.com>
Message-Id: <20220516154310.3685678-1-Ashish.Kalra@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agox86/fpu: KVM: Set the base guest FPU uABI size to sizeof(struct kvm_xsave)
Sean Christopherson [Wed, 4 May 2022 00:12:19 +0000 (00:12 +0000)]
x86/fpu: KVM: Set the base guest FPU uABI size to sizeof(struct kvm_xsave)

Set the starting uABI size of KVM's guest FPU to 'struct kvm_xsave',
i.e. to KVM's historical uABI size.  When saving FPU state for usersapce,
KVM (well, now the FPU) sets the FP+SSE bits in the XSAVE header even if
the host doesn't support XSAVE.  Setting the XSAVE header allows the VM
to be migrated to a host that does support XSAVE without the new host
having to handle FPU state that may or may not be compatible with XSAVE.

Setting the uABI size to the host's default size results in out-of-bounds
writes (setting the FP+SSE bits) and data corruption (that is thankfully
caught by KASAN) when running on hosts without XSAVE, e.g. on Core2 CPUs.

WARN if the default size is larger than KVM's historical uABI size; all
features that can push the FPU size beyond the historical size must be
opt-in.

  ==================================================================
  BUG: KASAN: slab-out-of-bounds in fpu_copy_uabi_to_guest_fpstate+0x86/0x130
  Read of size 8 at addr ffff888011e33a00 by task qemu-build/681
  CPU: 1 PID: 681 Comm: qemu-build Not tainted 5.18.0-rc5-KASAN-amd64 #1
  Hardware name:  /DG35EC, BIOS ECG3510M.86A.0118.2010.0113.1426 01/13/2010
  Call Trace:
   <TASK>
   dump_stack_lvl+0x34/0x45
   print_report.cold+0x45/0x575
   kasan_report+0x9b/0xd0
   fpu_copy_uabi_to_guest_fpstate+0x86/0x130
   kvm_arch_vcpu_ioctl+0x72a/0x1c50 [kvm]
   kvm_vcpu_ioctl+0x47f/0x7b0 [kvm]
   __x64_sys_ioctl+0x5de/0xc90
   do_syscall_64+0x31/0x50
   entry_SYSCALL_64_after_hwframe+0x44/0xae
   </TASK>
  Allocated by task 0:
  (stack is not available)
  The buggy address belongs to the object at ffff888011e33800
   which belongs to the cache kmalloc-512 of size 512
  The buggy address is located 0 bytes to the right of
   512-byte region [ffff888011e33800ffff888011e33a00)
  The buggy address belongs to the physical page:
  page:0000000089cd4adb refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x11e30
  head:0000000089cd4adb order:2 compound_mapcount:0 compound_pincount:0
  flags: 0x4000000000010200(slab|head|zone=1)
  raw: 4000000000010200 dead000000000100 dead000000000122 ffff888001041c80
  raw: 0000000000000000 0000000080100010 00000001ffffffff 0000000000000000
  page dumped because: kasan: bad access detected
  Memory state around the buggy address:
   ffff888011e33900: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   ffff888011e33980: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  >ffff888011e33a00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
                     ^
   ffff888011e33a80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
   ffff888011e33b00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
  ==================================================================
  Disabling lock debugging due to kernel taint

Fixes: be50b2065dfa ("kvm: x86: Add support for getting/setting expanded xstate buffer")
Fixes: c60427dd50ba ("x86/fpu: Add uabi_size to guest_fpu")
Reported-by: Zdenek Kaspar <zkaspar82@gmail.com>
Cc: Maciej S. Szmigiero <mail@maciej.szmigiero.name>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: kvm@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Tested-by: Zdenek Kaspar <zkaspar82@gmail.com>
Message-Id: <20220504001219.983513-1-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agos390/uv_uapi: depend on CONFIG_S390
Paolo Bonzini [Mon, 23 May 2022 19:23:21 +0000 (15:23 -0400)]
s390/uv_uapi: depend on CONFIG_S390

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoMerge tag 'kvm-s390-next-5.19-1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Paolo Bonzini [Wed, 25 May 2022 09:11:21 +0000 (05:11 -0400)]
Merge tag 'kvm-s390-next-5.19-1' of git://git./linux/kernel/git/kvms390/linux into HEAD

KVM: s390: Fix and feature for 5.19

- ultravisor communication device driver
- fix TEID on terminating storage key ops

2 years agoMerge tag 'kvm-riscv-5.19-1' of https://github.com/kvm-riscv/linux into HEAD
Paolo Bonzini [Wed, 25 May 2022 09:09:49 +0000 (05:09 -0400)]
Merge tag 'kvm-riscv-5.19-1' of https://github.com/kvm-riscv/linux into HEAD

KVM/riscv changes for 5.19

- Added Sv57x4 support for G-stage page table
- Added range based local HFENCE functions
- Added remote HFENCE functions based on VCPU requests
- Added ISA extension registers in ONE_REG interface
- Updated KVM RISC-V maintainers entry to cover selftests support

2 years agoMerge tag 'kvmarm-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmar...
Paolo Bonzini [Fri, 20 May 2022 11:16:27 +0000 (07:16 -0400)]
Merge tag 'kvmarm-5.19' of git://git./linux/kernel/git/kvmarm/kvmarm into HEAD

KVM/arm64 updates for 5.19

- Add support for the ARMv8.6 WFxT extension

- Guard pages for the EL2 stacks

- Trap and emulate AArch32 ID registers to hide unsupported features

- Ability to select and save/restore the set of hypercalls exposed
  to the guest

- Support for PSCI-initiated suspend in collaboration with userspace

- GICv3 register-based LPI invalidation support

- Move host PMU event merging into the vcpu data structure

- GICv3 ITS save/restore fixes

- The usual set of small-scale cleanups and fixes

[Due to the conflict, KVM_SYSTEM_EVENT_SEV_TERM is relocated
 from 4 to 6. - Paolo]

2 years agoKVM: selftests: x86: Fix test failure on arch lbr capable platforms
Yang Weijiang [Thu, 12 May 2022 08:40:46 +0000 (04:40 -0400)]
KVM: selftests: x86: Fix test failure on arch lbr capable platforms

On Arch LBR capable platforms, LBR_FMT in perf capability msr is 0x3f,
so the last format test will fail. Use a true invalid format(0x30) for
the test if it's running on these platforms. Opportunistically change
the file name to reflect the tests actually carried out.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Yang Weijiang <weijiang.yang@intel.com>
Message-Id: <20220512084046.105479-1-weijiang.yang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: LAPIC: Trace LAPIC timer expiration on every vmentry
Wanpeng Li [Tue, 26 Apr 2022 08:25:51 +0000 (01:25 -0700)]
KVM: LAPIC: Trace LAPIC timer expiration on every vmentry

In commit ec0671d5684a ("KVM: LAPIC: Delay trace_kvm_wait_lapic_expire
tracepoint to after vmexit", 2019-06-04), trace_kvm_wait_lapic_expire
was moved after guest_exit_irqoff() because invoking tracepoints within
kvm_guest_enter/kvm_guest_exit caused a lockdep splat.

These days this is not necessary, because commit 87fa7f3e98a1 ("x86/kvm:
Move context tracking where it belongs", 2020-07-09) restricted
the RCU extended quiescent state to be closer to vmentry/vmexit.
Moving the tracepoint back to __kvm_wait_lapic_expire is more accurate,
because it will be reported even if vcpu_enter_guest causes multiple
vmentries via the IPI/Timer fast paths, and it allows the removal of
advance_expire_delta.

Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
Message-Id: <1650961551-38390-1-git-send-email-wanpengli@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: s390: selftest: Test suppression indication on key prot exception
Janis Schoetterl-Glausch [Thu, 12 May 2022 13:10:18 +0000 (15:10 +0200)]
KVM: s390: selftest: Test suppression indication on key prot exception

Check that suppression is not indicated on injection of a key checked
protection exception caused by a memop after it already modified guest
memory, as that violates the definition of suppression.

Signed-off-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Link: https://lore.kernel.org/r/20220512131019.2594948-3-scgl@linux.ibm.com
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
2 years agoKVM: s390: Don't indicate suppression on dirtying, failing memop
Janis Schoetterl-Glausch [Thu, 12 May 2022 13:10:17 +0000 (15:10 +0200)]
KVM: s390: Don't indicate suppression on dirtying, failing memop

If user space uses a memop to emulate an instruction and that
memop fails, the execution of the instruction ends.
Instruction execution can end in different ways, one of which is
suppression, which requires that the instruction execute like a no-op.
A writing memop that spans multiple pages and fails due to key
protection may have modified guest memory, as a result, the likely
correct ending is termination. Therefore, do not indicate a
suppressing instruction ending in this case.

Signed-off-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/r/20220512131019.2594948-2-scgl@linux.ibm.com
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
2 years agoselftests: drivers/s390x: Add uvdevice tests
Steffen Eiden [Tue, 10 May 2022 14:47:24 +0000 (14:47 +0000)]
selftests: drivers/s390x: Add uvdevice tests

Adds some selftests to test ioctl error paths of the uv-uapi.
The Kconfig S390_UV_UAPI must be selected and the Ultravisor facility
must be available. The test can be executed by non-root, however, the
uvdevice special file /dev/uv must be accessible for reading and
writing which may imply root privileges.

  ./test-uv-device
  TAP version 13
  1..6
  # Starting 6 tests from 3 test cases.
  #  RUN           uvio_fixture.att.fault_ioctl_arg ...
  #            OK  uvio_fixture.att.fault_ioctl_arg
  ok 1 uvio_fixture.att.fault_ioctl_arg
  #  RUN           uvio_fixture.att.fault_uvio_arg ...
  #            OK  uvio_fixture.att.fault_uvio_arg
  ok 2 uvio_fixture.att.fault_uvio_arg
  #  RUN           uvio_fixture.att.inval_ioctl_cb ...
  #            OK  uvio_fixture.att.inval_ioctl_cb
  ok 3 uvio_fixture.att.inval_ioctl_cb
  #  RUN           uvio_fixture.att.inval_ioctl_cmd ...
  #            OK  uvio_fixture.att.inval_ioctl_cmd
  ok 4 uvio_fixture.att.inval_ioctl_cmd
  #  RUN           attest_fixture.att_inval_request ...
  #            OK  attest_fixture.att_inval_request
  ok 5 attest_fixture.att_inval_request
  #  RUN           attest_fixture.att_inval_addr ...
  #            OK  attest_fixture.att_inval_addr
  ok 6 attest_fixture.att_inval_addr
  # PASSED: 6 / 6 tests passed.
  # Totals: pass:6 fail:0 xfail:0 xpass:0 skip:0 error:0

Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Acked-by: Janosch Frank <frankja@linux.ibm.com>
Message-Id: <20220510144724.3321985-3-seiden@linux.ibm.com>
Link: https://lore.kernel.org/kvm/20220510144724.3321985-3-seiden@linux.ibm.com/
Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
2 years agodrivers/s390/char: Add Ultravisor io device
Steffen Eiden [Mon, 16 May 2022 11:33:35 +0000 (11:33 +0000)]
drivers/s390/char: Add Ultravisor io device

This patch adds a new miscdevice to expose some Ultravisor functions
to userspace. Userspace can send IOCTLs to the uvdevice that will then
emit a corresponding Ultravisor Call and hands the result over to
userspace. The uvdevice is available if the Ultravisor Call facility is
present.
Userspace can call the Retrieve Attestation Measurement
Ultravisor Call using IOCTLs on the uvdevice.

The uvdevice will do some sanity checks first.
Then, copy the request data to kernel space, build the UVCB,
perform the UV call, and copy the result back to userspace.

Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/kvm/20220516113335.338212-1-seiden@linux.ibm.com/
Message-Id: <20220516113335.338212-1-seiden@linux.ibm.com>
Signed-off-by: Janosch Frank <frankja@linux.ibm.com> (whitespace and tristate fixes, pick)
2 years agoMAINTAINERS: Update KVM RISC-V entry to cover selftests support
Anup Patel [Mon, 9 May 2022 05:53:16 +0000 (11:23 +0530)]
MAINTAINERS: Update KVM RISC-V entry to cover selftests support

We update KVM RISC-V maintainers entry to include appropriate KVM
selftests directories so that RISC-V related KVM selftests patches
are CC'ed to KVM RISC-V mailing list.

Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Introduce ISA extension register
Atish Patra [Mon, 9 May 2022 18:29:37 +0000 (11:29 -0700)]
RISC-V: KVM: Introduce ISA extension register

Currently, there is no provision for vmm (qemu-kvm or kvmtool) to
query about multiple-letter ISA extensions. The config register
is only used for base single letter ISA extensions.

A new ISA extension register is added that will allow the vmm
to query about any ISA extension one at a time. It is enabled for
both single letter or multi-letter ISA extensions. The ISA extension
register is useful to if the vmm requires to retrieve/set single
extension while the config register should be used if all the base
ISA extension required to retrieve or set.

For any multi-letter ISA extensions, the new register interface
must be used.

Signed-off-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Cleanup stale TLB entries when host CPU changes
Anup Patel [Mon, 9 May 2022 05:14:11 +0000 (10:44 +0530)]
RISC-V: KVM: Cleanup stale TLB entries when host CPU changes

On RISC-V platforms with hardware VMID support, we share same
VMID for all VCPUs of a particular Guest/VM. This means we might
have stale G-stage TLB entries on the current Host CPU due to
some other VCPU of the same Guest which ran previously on the
current Host CPU.

To cleanup stale TLB entries, we simply flush all G-stage TLB
entries by VMID whenever underlying Host CPU changes for a VCPU.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Add remote HFENCE functions based on VCPU requests
Anup Patel [Mon, 9 May 2022 05:14:05 +0000 (10:44 +0530)]
RISC-V: KVM: Add remote HFENCE functions based on VCPU requests

The generic KVM has support for VCPU requests which can be used
to do arch-specific work in the run-loop. We introduce remote
HFENCE functions which will internally use VCPU requests instead
of host SBI calls.

Advantages of doing remote HFENCEs as VCPU requests are:
1) Multiple VCPUs of a Guest may be running on different Host CPUs
   so it is not always possible to determine the Host CPU mask for
   doing Host SBI call. For example, when VCPU X wants to do HFENCE
   on VCPU Y, it is possible that VCPU Y is blocked or in user-space
   (i.e. vcpu->cpu < 0).
2) To support nested virtualization, we will be having a separate
   shadow G-stage for each VCPU and a common host G-stage for the
   entire Guest/VM. The VCPU requests based remote HFENCEs helps
   us easily synchronize the common host G-stage and shadow G-stage
   of each VCPU without any additional IPI calls.

This is also a preparatory patch for upcoming nested virtualization
support where we will be having a shadow G-stage page table for
each Guest VCPU.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Reduce KVM_MAX_VCPUS value
Anup Patel [Mon, 9 May 2022 05:14:00 +0000 (10:44 +0530)]
RISC-V: KVM: Reduce KVM_MAX_VCPUS value

Currently, the KVM_MAX_VCPUS value is 16384 for RV64 and 128
for RV32.

The KVM_MAX_VCPUS value is too high for RV64 and too low for
RV32 compared to other architectures (e.g. x86 sets it to 1024
and ARM64 sets it to 512). The too high value of KVM_MAX_VCPUS
on RV64 also leads to VCPU mask on stack consuming 2KB.

We set KVM_MAX_VCPUS to 1024 for both RV64 and RV32 to be
aligned other architectures.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Introduce range based local HFENCE functions
Anup Patel [Mon, 9 May 2022 05:13:52 +0000 (10:43 +0530)]
RISC-V: KVM: Introduce range based local HFENCE functions

Various  __kvm_riscv_hfence_xyz() functions implemented in the
kvm/tlb.S are equivalent to corresponding HFENCE.GVMA instructions
and we don't have range based local HFENCE functions.

This patch provides complete set of local HFENCE functions which
supports range based TLB invalidation and supports HFENCE.VVMA
based functions. This is also a preparatory patch for upcoming
Svinval support in KVM RISC-V.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Treat SBI HFENCE calls as NOPs
Anup Patel [Mon, 9 May 2022 05:13:46 +0000 (10:43 +0530)]
RISC-V: KVM: Treat SBI HFENCE calls as NOPs

We should treat SBI HFENCE calls as NOPs until nested virtualization
is supported by KVM RISC-V. This will help us test booting a hypervisor
under KVM RISC-V.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Add Sv57x4 mode support for G-stage
Anup Patel [Mon, 9 May 2022 05:13:39 +0000 (10:43 +0530)]
RISC-V: KVM: Add Sv57x4 mode support for G-stage

Latest QEMU supports G-stage Sv57x4 mode so this patch extends KVM
RISC-V G-stage handling to detect and use Sv57x4 mode when available.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoRISC-V: KVM: Use G-stage name for hypervisor page table
Anup Patel [Mon, 9 May 2022 05:13:30 +0000 (10:43 +0530)]
RISC-V: KVM: Use G-stage name for hypervisor page table

The two-stage address translation defined by the RISC-V privileged
specification defines: VS-stage (guest virtual address to guest
physical address) programmed by the Guest OS  and G-stage (guest
physical addree to host physical address) programmed by the
hypervisor.

To align with above terminology, we replace "stage2" with "gstage"
and "Stage2" with "G-stage" name everywhere in KVM RISC-V sources.

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoKVM: selftests: riscv: Remove unneeded semicolon
Jiapeng Chong [Fri, 6 May 2022 09:45:12 +0000 (17:45 +0800)]
KVM: selftests: riscv: Remove unneeded semicolon

Fix the following coccicheck warnings:

./tools/testing/selftests/kvm/lib/riscv/processor.c:353:3-4: Unneeded
semicolon.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoKVM: selftests: riscv: Improve unexpected guest trap handling
Anup Patel [Sat, 9 Apr 2022 04:32:36 +0000 (10:02 +0530)]
KVM: selftests: riscv: Improve unexpected guest trap handling

Currently, we simply hang using "while (1) ;" upon any unexpected
guest traps because the default guest trap handler is guest_hang().

The above approach is not useful to anyone because KVM selftests
users will only see a hung application upon any unexpected guest
trap.

This patch improves unexpected guest trap handling for KVM RISC-V
selftests by doing the following:
1) Return to host user-space
2) Dump VCPU registers
3) Die using TEST_ASSERT(0, ...)

Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Tested-by: Mayuresh Chitale <mchitale@ventanamicro.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
2 years agoMerge branch kvm-arm64/its-save-restore-fixes-5.19 into kvmarm-master/next
Marc Zyngier [Mon, 16 May 2022 16:48:36 +0000 (17:48 +0100)]
Merge branch kvm-arm64/its-save-restore-fixes-5.19 into kvmarm-master/next

* kvm-arm64/its-save-restore-fixes-5.19:
  : .
  : Tighten the ITS save/restore infrastructure to fail early rather
  : than late. Patches courtesy of Rocardo Koller.
  : .
  KVM: arm64: vgic: Undo work in failed ITS restores
  KVM: arm64: vgic: Do not ignore vgic_its_restore_cte failures
  KVM: arm64: vgic: Add more checks when restoring ITS tables
  KVM: arm64: vgic: Check that new ITEs could be saved in guest memory

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoMerge branch kvm-arm64/misc-5.19 into kvmarm-master/next
Marc Zyngier [Mon, 16 May 2022 16:48:36 +0000 (17:48 +0100)]
Merge branch kvm-arm64/misc-5.19 into kvmarm-master/next

* kvm-arm64/misc-5.19:
  : .
  : Misc fixes and general improvements for KVMM/arm64:
  :
  : - Better handle out of sequence sysregs in the global tables
  :
  : - Remove a couple of unnecessary loads from constant pool
  :
  : - Drop unnecessary pKVM checks
  :
  : - Add all known M1 implementations to the SEIS workaround
  :
  : - Cleanup kerneldoc warnings
  : .
  KVM: arm64: vgic-v3: List M1 Pro/Max as requiring the SEIS workaround
  KVM: arm64: pkvm: Don't mask already zeroed FEAT_SVE
  KVM: arm64: pkvm: Drop unnecessary FP/SIMD trap handler
  KVM: arm64: nvhe: Eliminate kernel-doc warnings
  KVM: arm64: Avoid unnecessary absolute addressing via literals
  KVM: arm64: Print emulated register table name when it is unsorted
  KVM: arm64: Don't BUG_ON() if emulated register table is unsorted

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoMerge branch kvm-arm64/per-vcpu-host-pmu-data into kvmarm-master/next
Marc Zyngier [Mon, 16 May 2022 16:48:36 +0000 (17:48 +0100)]
Merge branch kvm-arm64/per-vcpu-host-pmu-data into kvmarm-master/next

* kvm-arm64/per-vcpu-host-pmu-data:
  : .
  : Pass the host PMU state in the vcpu to avoid the use of additional
  : shared memory between EL1 and EL2 (this obviously only applies
  : to nVHE and Protected setups).
  :
  : Patches courtesy of Fuad Tabba.
  : .
  KVM: arm64: pmu: Restore compilation when HW_PERF_EVENTS isn't selected
  KVM: arm64: Reenable pmu in Protected Mode
  KVM: arm64: Pass pmu events to hyp via vcpu
  KVM: arm64: Repack struct kvm_pmu to reduce size
  KVM: arm64: Wrapper for getting pmu_events

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoMerge branch kvm-arm64/vgic-invlpir into kvmarm-master/next
Marc Zyngier [Mon, 16 May 2022 16:48:35 +0000 (17:48 +0100)]
Merge branch kvm-arm64/vgic-invlpir into kvmarm-master/next

* kvm-arm64/vgic-invlpir:
  : .
  : Implement MMIO-based LPI invalidation for vGICv3.
  : .
  KVM: arm64: vgic-v3: Advertise GICR_CTLR.{IR, CES} as a new GICD_IIDR revision
  KVM: arm64: vgic-v3: Implement MMIO-based LPI invalidation
  KVM: arm64: vgic-v3: Expose GICR_CTLR.RWP when disabling LPIs
  irqchip/gic-v3: Exposes bit values for GICR_CTLR.{IR, CES}

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoMerge branch kvm-arm64/psci-suspend into kvmarm-master/next
Marc Zyngier [Mon, 16 May 2022 16:48:20 +0000 (17:48 +0100)]
Merge branch kvm-arm64/psci-suspend into kvmarm-master/next

* kvm-arm64/psci-suspend:
  : .
  : Add support for PSCI SYSTEM_SUSPEND and allow userspace to
  : filter the wake-up events.
  :
  : Patches courtesy of Oliver.
  : .
  Documentation: KVM: Fix title level for PSCI_SUSPEND
  selftests: KVM: Test SYSTEM_SUSPEND PSCI call
  selftests: KVM: Refactor psci_test to make it amenable to new tests
  selftests: KVM: Use KVM_SET_MP_STATE to power off vCPU in psci_test
  selftests: KVM: Create helper for making SMCCC calls
  selftests: KVM: Rename psci_cpu_on_test to psci_test
  KVM: arm64: Implement PSCI SYSTEM_SUSPEND
  KVM: arm64: Add support for userspace to suspend a vCPU
  KVM: arm64: Return a value from check_vcpu_requests()
  KVM: arm64: Rename the KVM_REQ_SLEEP handler
  KVM: arm64: Track vCPU power state using MP state values
  KVM: arm64: Dedupe vCPU power off helpers
  KVM: arm64: Don't depend on fallthrough to hide SYSTEM_RESET2

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoMerge branch kvm-arm64/hcall-selection into kvmarm-master/next
Marc Zyngier [Mon, 16 May 2022 16:47:03 +0000 (17:47 +0100)]
Merge branch kvm-arm64/hcall-selection into kvmarm-master/next

* kvm-arm64/hcall-selection:
  : .
  : Introduce a new set of virtual sysregs for userspace to
  : select the hypercalls it wants to see exposed to the guest.
  :
  : Patches courtesy of Raghavendra and Oliver.
  : .
  KVM: arm64: Fix hypercall bitmap writeback when vcpus have already run
  KVM: arm64: Hide KVM_REG_ARM_*_BMAP_BIT_COUNT from userspace
  Documentation: Fix index.rst after psci.rst renaming
  selftests: KVM: aarch64: Add the bitmap firmware registers to get-reg-list
  selftests: KVM: aarch64: Introduce hypercall ABI test
  selftests: KVM: Create helper for making SMCCC calls
  selftests: KVM: Rename psci_cpu_on_test to psci_test
  tools: Import ARM SMCCC definitions
  Docs: KVM: Add doc for the bitmap firmware registers
  Docs: KVM: Rename psci.rst to hypercalls.rst
  KVM: arm64: Add vendor hypervisor firmware register
  KVM: arm64: Add standard hypervisor firmware register
  KVM: arm64: Setup a framework for hypercall bitmap firmware registers
  KVM: arm64: Factor out firmware register handling from psci.c

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoKVM: arm64: Fix hypercall bitmap writeback when vcpus have already run
Marc Zyngier [Mon, 16 May 2022 16:32:54 +0000 (17:32 +0100)]
KVM: arm64: Fix hypercall bitmap writeback when vcpus have already run

We generally want to disallow hypercall bitmaps being changed
once vcpus have already run. But we must allow the write if
the written value is unchanged so that userspace can rewrite
the register file on reboot, for example.

Without this, a QEMU-based VM will fail to reboot correctly.

The original code was correct, and it is me that introduced
the regression.

Fixes: 05714cab7d63 ("KVM: arm64: Setup a framework for hypercall bitmap firmware registers")
Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoKVM: arm64: vgic: Undo work in failed ITS restores
Ricardo Koller [Tue, 10 May 2022 00:16:33 +0000 (17:16 -0700)]
KVM: arm64: vgic: Undo work in failed ITS restores

Failed ITS restores should clean up all state restored until the
failure. There is some cleanup already present when failing to restore
some tables, but it's not complete. Add the missing cleanup.

Note that this changes the behavior in case of a failed restore of the
device tables.

restore ioctl:
1. restore collection tables
2. restore device tables

With this commit, failures in 2. clean up everything created so far,
including state created by 1.

Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510001633.552496-5-ricarkol@google.com
2 years agoKVM: arm64: vgic: Do not ignore vgic_its_restore_cte failures
Ricardo Koller [Tue, 10 May 2022 00:16:32 +0000 (17:16 -0700)]
KVM: arm64: vgic: Do not ignore vgic_its_restore_cte failures

Restoring a corrupted collection entry (like an out of range ID) is
being ignored and treated as success. More specifically, a
vgic_its_restore_cte failure is treated as success by
vgic_its_restore_collection_table.  vgic_its_restore_cte uses positive
and negative numbers to return error, and +1 to return success.  The
caller then uses "ret > 0" to check for success.

Fix this by having vgic_its_restore_cte only return negative numbers on
error.  Do this by changing alloc_collection return codes to only return
negative numbers on error.

Signed-off-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510001633.552496-4-ricarkol@google.com
2 years agoKVM: arm64: vgic: Add more checks when restoring ITS tables
Ricardo Koller [Tue, 10 May 2022 00:16:31 +0000 (17:16 -0700)]
KVM: arm64: vgic: Add more checks when restoring ITS tables

Try to improve the predictability of ITS save/restores (and debuggability
of failed ITS saves) by failing early on restore when trying to read
corrupted tables.

Restoring the ITS tables does some checks for corrupted tables, but not as
many as in a save: an overflowing device ID will be detected on save but
not on restore.  The consequence is that restoring a corrupted table won't
be detected until the next save; including the ITS not working as expected
after the restore.  As an example, if the guest sets tables overlapping
each other, which would most likely result in some corrupted table, this is
what we would see from the host point of view:

guest sets base addresses that overlap each other
save ioctl
restore ioctl
save ioctl (fails)

Ideally, we would like the first save to fail, but overlapping tables could
actually be intended by the guest. So, let's at least fail on the restore
with some checks: like checking that device and event IDs don't overflow
their tables.

Signed-off-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510001633.552496-3-ricarkol@google.com
2 years agoKVM: arm64: vgic: Check that new ITEs could be saved in guest memory
Ricardo Koller [Tue, 10 May 2022 00:16:30 +0000 (17:16 -0700)]
KVM: arm64: vgic: Check that new ITEs could be saved in guest memory

Try to improve the predictability of ITS save/restores by failing
commands that would lead to failed saves. More specifically, fail any
command that adds an entry into an ITS table that is not in guest
memory, which would otherwise lead to a failed ITS save ioctl. There
are already checks for collection and device entries, but not for
ITEs.  Add the corresponding check for the ITT when adding ITEs.

Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Ricardo Koller <ricarkol@google.com>
Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510001633.552496-2-ricarkol@google.com
2 years agoKVM: arm64: pmu: Restore compilation when HW_PERF_EVENTS isn't selected
Marc Zyngier [Mon, 16 May 2022 12:02:24 +0000 (13:02 +0100)]
KVM: arm64: pmu: Restore compilation when HW_PERF_EVENTS isn't selected

Moving kvm_pmu_events into the vcpu (and refering to it) broke the
somewhat unusual case where the kernel has no support for a PMU
at all.

In order to solve this, move things around a bit so that we can
easily avoid refering to the pmu structure outside of PMU-aware
code. As a bonus, pmu.c isn't compiled in when HW_PERF_EVENTS
isn't selected.

Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Fuad Tabba <tabba@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/202205161814.KQHpOzsJ-lkp@intel.com
2 years agoLinux 5.18-rc7
Linus Torvalds [Mon, 16 May 2022 01:08:58 +0000 (18:08 -0700)]
Linux 5.18-rc7

2 years agoMerge tag 'driver-core-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 15 May 2022 15:08:51 +0000 (08:08 -0700)]
Merge tag 'driver-core-5.18-rc7' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fixes from Greg KH:
 "Here is one fix, and three documentation updates for 5.18-rc7.

  The fix is for the firmware loader which resolves a long-reported
  problem where the credentials of the firmware loader could be set to a
  userspace process without enough permissions to actually load the
  firmware image. Many Android vendors have been reporting this for
  quite some time.

  The documentation updates are for the embargoed-hardware-issues.rst
  file to add a new entry, change an existing one, and sort the list to
  make changes easier in the future.

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'driver-core-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  Documentation/process: Update ARM contact for embargoed hardware issues
  Documentation/process: Add embargoed HW contact for Ampere Computing
  Documentation/process: Make groups alphabetical and use tabs consistently
  firmware_loader: use kernel credentials when reading firmware

2 years agoMerge tag 'char-misc-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Sun, 15 May 2022 15:07:07 +0000 (08:07 -0700)]
Merge tag 'char-misc-5.18-rc7' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are two small driver fixes for 5.18-rc7 that resolve reported
  problems:

   - slimbus driver irq bugfix

   - interconnect sync state bugfix

  Both of these have been in linux-next with no reported problems"

* tag 'char-misc-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  slimbus: qcom: Fix IRQ check in qcom_slim_probe
  interconnect: Restore sync state by ignoring ipa-virt in provider count

2 years agoMerge tag 'tty-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sun, 15 May 2022 15:05:04 +0000 (08:05 -0700)]
Merge tag 'tty-5.18-rc7' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial driver fixes from Greg KH:
 "Here are some small tty n_gsm and serial driver fixes for 5.18-rc7
  that resolve reported problems. They include:

   - n_gsm fixes for reported issues

   - 8250_mtk driver fixes for some platforms

   - fsl_lpuart driver fix for reported problem.

   - digicolor driver fix for reported problem.

  All have been in linux-next for a while with no reported problems"

* tag 'tty-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  fsl_lpuart: Don't enable interrupts too early
  tty: n_gsm: fix invalid gsmtty_write_room() result
  tty: n_gsm: fix mux activation issues in gsm_config()
  tty: n_gsm: fix buffer over-read in gsm_dlci_data()
  serial: 8250_mtk: Fix register address for XON/XOFF character
  serial: 8250_mtk: Make sure to select the right FEATURE_SEL
  serial: 8250_mtk: Fix UART_EFR register address
  tty/serial: digicolor: fix possible null-ptr-deref in digicolor_uart_probe()

2 years agoMerge tag 'usb-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 15 May 2022 15:03:24 +0000 (08:03 -0700)]
Merge tag 'usb-5.18-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small fixes for reported issues with some USB drivers.
  They include:

   - xhci fixes for xhci-mtk platform driver

   - typec driver fixes for reported problems.

   - cdc-wdm read-stuck fix

   - gadget driver fix for reported race condition

   - new usb-serial driver ids

  All of these have been in linux-next with no reported problems"

* tag 'usb-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: xhci-mtk: remove bandwidth budget table
  usb: xhci-mtk: fix fs isoc's transfer error
  usb: gadget: fix race when gadget driver register via ioctl
  usb: typec: tcpci_mt6360: Update for BMC PHY setting
  usb: gadget: uvc: allow for application to cleanly shutdown
  usb: typec: tcpci: Don't skip cleanup in .remove() on error
  usb: cdc-wdm: fix reading stuck on device close
  USB: serial: qcserial: add support for Sierra Wireless EM7590
  USB: serial: option: add Fibocom MA510 modem
  USB: serial: option: add Fibocom L610 modem
  USB: serial: pl2303: add device id for HP LM930 Display

2 years agoMerge tag 'powerpc-5.18-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sun, 15 May 2022 13:46:03 +0000 (06:46 -0700)]
Merge tag 'powerpc-5.18-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fix from Michael Ellerman:

 - Fix KVM PR on 32-bit, which was broken by some MMU code refactoring.

Thanks to: Alexander Graf, and Matt Evans.

* tag 'powerpc-5.18-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  KVM: PPC: Book3S PR: Enable MSR_DR for switch_mmu_context()

2 years agoMerge tag 'x86-urgent-2022-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 15 May 2022 13:42:40 +0000 (06:42 -0700)]
Merge tag 'x86-urgent-2022-05-15' of git://git./linux/kernel/git/tip/tip

Pull x86 fix from Thomas Gleixner:
 "A single fix for the handling of unpopulated sub-pmd spaces.

  The copy & pasta from the corresponding s390 code screwed up the
  address calculation for marking the sub-pmd ranges via memset by
  omitting the ALIGN_DOWN() to calculate the proper start address.

  It's a mystery why this code is not generic and shared because there
  is nothing architecture specific in there, but that's too intrusive
  for a backportable fix"

* tag 'x86-urgent-2022-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm: Fix marking of unused sub-pmd ranges

2 years agoMerge tag 'sched-urgent-2022-05-15' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 15 May 2022 13:40:11 +0000 (06:40 -0700)]
Merge tag 'sched-urgent-2022-05-15' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Thomas Gleixner:
 "The recent expansion of the sched switch tracepoint inserted a new
  argument in the middle of the arguments. This reordering broke BPF
  programs which relied on the old argument list.

  While tracepoints are not considered stable ABI, it's not trivial to
  make BPF cope with such a change, but it's being worked on. For now
  restore the original argument order and move the new argument to the
  end of the argument list"

* tag 'sched-urgent-2022-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/tracing: Append prev_state to tp args instead

2 years agoMerge tag 'irq-urgent-2022-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 15 May 2022 13:37:05 +0000 (06:37 -0700)]
Merge tag 'irq-urgent-2022-05-15' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Thomas Gleixner:
 "A single fix for a recent (introduced in 5.16) regression in the core
  interrupt code.

  The consolidation of the interrupt handler invocation code added an
  unconditional warning when generic_handle_domain_irq() is invoked from
  outside hard interrupt context. That's overbroad as the requirement
  for invoking these handlers in hard interrupt context is only required
  for certain interrupt types. The subsequently called code already
  contains a warning which triggers conditionally for interrupt chips
  which indicate this requirement in their properties.

  Remove the overbroad one"

* tag 'irq-urgent-2022-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Remove WARN_ON_ONCE() in generic_handle_domain_irq()

2 years agoKVM: arm64: Hide KVM_REG_ARM_*_BMAP_BIT_COUNT from userspace
Marc Zyngier [Sun, 15 May 2022 10:36:24 +0000 (11:36 +0100)]
KVM: arm64: Hide KVM_REG_ARM_*_BMAP_BIT_COUNT from userspace

These constants will change over time, and userspace has no
business knowing about them. Hide them behind __KERNEL__.

Signed-off-by: Marc Zyngier <maz@kernel.org>
2 years agoKVM: arm64: Reenable pmu in Protected Mode
Fuad Tabba [Tue, 10 May 2022 09:57:10 +0000 (09:57 +0000)]
KVM: arm64: Reenable pmu in Protected Mode

Now that the pmu code does not access hyp data, reenable it in
protected mode.

Once fully supported, protected VMs will not have pmu support,
since that could leak information. However, non-protected VMs in
protected mode should have pmu support if available.

Signed-off-by: Fuad Tabba <tabba@google.com>
Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510095710.148178-5-tabba@google.com
2 years agoKVM: arm64: Pass pmu events to hyp via vcpu
Fuad Tabba [Tue, 10 May 2022 09:57:09 +0000 (09:57 +0000)]
KVM: arm64: Pass pmu events to hyp via vcpu

Instead of the host accessing hyp data directly, pass the pmu
events of the current cpu to hyp via the vcpu.

This adds 64 bits (in two fields) to the vcpu that need to be
synced before every vcpu run in nvhe and protected modes.
However, it isolates the hypervisor from the host, which allows
us to use pmu in protected mode in a subsequent patch.

No visible side effects in behavior intended.

Signed-off-by: Fuad Tabba <tabba@google.com>
Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510095710.148178-4-tabba@google.com
2 years agoKVM: arm64: Repack struct kvm_pmu to reduce size
Fuad Tabba [Tue, 10 May 2022 09:57:08 +0000 (09:57 +0000)]
KVM: arm64: Repack struct kvm_pmu to reduce size

struct kvm_pmu has 2 holes using 10 bytes. This is instantiated
in all vcpus, so it adds up. Repack the structures to remove the
holes.

No functional change intended.

Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Fuad Tabba <tabba@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510095710.148178-3-tabba@google.com
2 years agoKVM: arm64: Wrapper for getting pmu_events
Fuad Tabba [Tue, 10 May 2022 09:57:07 +0000 (09:57 +0000)]
KVM: arm64: Wrapper for getting pmu_events

Eases migrating away from using hyp data and simplifies the code.

No functional change intended.

Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Fuad Tabba <tabba@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220510095710.148178-2-tabba@google.com
2 years agoKVM: arm64: vgic-v3: List M1 Pro/Max as requiring the SEIS workaround
Marc Zyngier [Sat, 14 May 2022 10:25:24 +0000 (11:25 +0100)]
KVM: arm64: vgic-v3: List M1 Pro/Max as requiring the SEIS workaround

Unsusprisingly, Apple M1 Pro/Max have the exact same defect as the
original M1 and generate random SErrors in the host when a guest
tickles the GICv3 CPU interface the wrong way.

Add the part numbers for both the CPU types found in these two
new implementations, and add them to the hall of shame. This also
applies to the Ultra version, as it is composed of 2 Max SoCs.

Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20220514102524.3188730-1-maz@kernel.org
2 years agoMerge tag 'perf-tools-fixes-for-v5.18-2022-05-14' of git://git.kernel.org/pub/scm...
Linus Torvalds [Sat, 14 May 2022 18:43:47 +0000 (11:43 -0700)]
Merge tag 'perf-tools-fixes-for-v5.18-2022-05-14' of git://git./linux/kernel/git/acme/linux

Pull perf tools fixes from Arnaldo Carvalho de Melo:

 - Fix two NDEBUG warnings in 'perf bench numa'

 - Fix ARM coresight `perf test` failure

 - Sync linux/kvm.h with the kernel sources

 - Add James and Mike as Arm64 performance events reviewers

* tag 'perf-tools-fixes-for-v5.18-2022-05-14' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  MAINTAINERS: Add James and Mike as Arm64 performance events reviewers
  tools headers UAPI: Sync linux/kvm.h with the kernel sources
  perf tests: Fix coresight `perf test` failure.
  perf bench: Fix two numa NDEBUG warnings

2 years agoMerge tag 'drm-fixes-2022-05-14' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 13 May 2022 23:20:25 +0000 (16:20 -0700)]
Merge tag 'drm-fixes-2022-05-14' of git://anongit.freedesktop.org/drm/drm

Pull more drm fixes from Dave Airlie:
 "Turns out I was right, some fixes hadn't made it to me yet. The vmwgfx
  ones also popped up later, but all seem like bad enough things to fix.
  The dma-buf, vc4 and nouveau ones are all pretty small.

  The fbdev fixes are a bit more complicated: a fix to cleanup fbdev
  devices properly, uncovered some use-after-free bugs in existing
  drivers. Then the fix for those bugs wasn't correct. This reverts that
  fix, and puts the proper fixes in place in the drivers to avoid the
  use-after-frees.

  This has had a fair number of eyes on it at this stage, and I'm
  confident enough that it puts things in the right place, and is less
  dangerous than reverting our way out of the initial change at this
  stage.

  fbdev:
   - revert NULL deref fix that turned into a use-after-free
   - prevent use-after-free in fbdev
   - efifb/simplefb/vesafb: fix cleanup paths to avoid use-after-frees

  dma-buf:
   - fix panic in stats setup

  vc4:
   - fix hdmi build

  nouveau:
   - tegra iommu present fix
   - fix leak in backlight name

  vmwgfx:
   - Black screen due to fences using FIFO checks on SVGA3
   - Random black screens on boot due to uninitialized drm_mode_fb_cmd2
   - Hangs on SVGA3 due to command buffers being used with gbobjects"

* tag 'drm-fixes-2022-05-14' of git://anongit.freedesktop.org/drm/drm:
  drm/vmwgfx: Disable command buffers on svga3 without gbobjects
  drm/vmwgfx: Initialize drm_mode_fb_cmd2
  drm/vmwgfx: Fix fencing on SVGAv3
  drm/vc4: hdmi: Fix build error for implicit function declaration
  dma-buf: call dma_buf_stats_setup after dmabuf is in valid list
  fbdev: efifb: Fix a use-after-free due early fb_info cleanup
  drm/nouveau: Fix a potential theorical leak in nouveau_get_backlight_name()
  drm/nouveau/tegra: Stop using iommu_present()
  fbdev: vesafb: Cleanup fb_info in .fb_destroy rather than .remove
  fbdev: efifb: Cleanup fb_info in .fb_destroy rather than .remove
  fbdev: simplefb: Cleanup fb_info in .fb_destroy rather than .remove
  fbdev: Prevent possible use-after-free in fb_release()
  Revert "fbdev: Make fb_release() return -ENODEV if fbdev was unregistered"

2 years agoMerge tag 'drm-misc-fixes-2022-05-13' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Fri, 13 May 2022 22:34:01 +0000 (08:34 +1000)]
Merge tag 'drm-misc-fixes-2022-05-13' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

Multiple fixes to fbdev to address a regression at unregistration, an
iommu detection improvement for nouveau, a memory leak fix for nouveau,
pointer dereference fix for dma_buf_file_release(), and a build breakage
fix for vc4

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20220513073044.ymayac7x7bzatrt7@houat
2 years agoMerge tag 'vmwgfx-drm-fixes-5.18-2022-05-13' of https://gitlab.freedesktop.org/zack...
Dave Airlie [Fri, 13 May 2022 22:29:41 +0000 (08:29 +1000)]
Merge tag 'vmwgfx-drm-fixes-5.18-2022-05-13' of https://gitlab.freedesktop.org/zack/vmwgfx into drm-fixes

vmwgfx fixes for:
- Black screen due to fences using FIFO checks on SVGA3
- Random black screens on boot due to uninitialized drm_mode_fb_cmd2
- Hangs on SVGA3 due to command buffers being used with gbobjects

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Zack Rusin <zackr@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/a1d32799e4c74b8540216376d7576bb783ca07ba.camel@vmware.com
2 years agoMerge tag 'gfs2-v5.18-rc4-fix3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 13 May 2022 21:32:53 +0000 (14:32 -0700)]
Merge tag 'gfs2-v5.18-rc4-fix3' of git://git./linux/kernel/git/gfs2/linux-gfs2

Pull gfs2 fixes from Andreas Gruenbacher:
 "We've finally identified commit dc732906c245 ("gfs2: Introduce flag
  for glock holder auto-demotion") to be the other cause of the
  filesystem corruption we've been seeing. This feature isn't strictly
  necessary anymore, so we've decided to stop using it for now.

  With this and the gfs_iomap_end rounding fix you've already seen
  ("gfs2: Fix filesystem block deallocation for short writes" in this
  pull request), we're corruption free again now.

   - Fix filesystem block deallocation for short writes.

   - Stop using glock holder auto-demotion for now.

   - Get rid of buffered writes inefficiencies due to page faults being
     disabled.

   - Minor other cleanups"

* tag 'gfs2-v5.18-rc4-fix3' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
  gfs2: Stop using glock holder auto-demotion for now
  gfs2: buffered write prefaulting
  gfs2: Align read and write chunks to the page cache
  gfs2: Pull return value test out of should_fault_in_pages
  gfs2: Clean up use of fault_in_iov_iter_{read,write}able
  gfs2: Variable rename
  gfs2: Fix filesystem block deallocation for short writes

2 years agogfs2: Stop using glock holder auto-demotion for now
Andreas Gruenbacher [Wed, 11 May 2022 16:27:12 +0000 (18:27 +0200)]
gfs2: Stop using glock holder auto-demotion for now

We're having unresolved issues with the glock holder auto-demotion mechanism
introduced in commit dc732906c245.  This mechanism was assumed to be essential
for avoiding frequent short reads and writes until commit 296abc0d91d8
("gfs2: No short reads or writes upon glock contention").  Since then,
when the inode glock is lost, it is simply re-acquired and the operation
is resumed.  This means that apart from the performance penalty, we
might as well drop the inode glock before faulting in pages, and
re-acquire it afterwards.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agogfs2: buffered write prefaulting
Andreas Gruenbacher [Wed, 4 May 2022 21:37:30 +0000 (23:37 +0200)]
gfs2: buffered write prefaulting

In gfs2_file_buffered_write, to increase the likelihood that all the
user memory we're trying to write will be resident in memory, carry out
the write in chunks and fault in each chunk of user memory before trying
to write it.  Otherwise, some workloads will trigger frequent short
"internal" writes, causing filesystem blocks to be allocated and then
partially deallocated again when writing into holes, which is wasteful
and breaks reservations.

Neither the chunked writes nor any of the short "internal" writes are
user visible.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Fri, 13 May 2022 20:13:48 +0000 (13:13 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Four fixes, all in drivers.

  These patches mosly fix error legs and exceptional conditions
  (scsi_dh_alua, qla2xxx). The lpfc fixes are for coding issues with
  lpfc features"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: lpfc: Correct BDE DMA address assignment for GEN_REQ_WQE
  scsi: lpfc: Fix split code for FLOGI on FCoE
  scsi: qla2xxx: Fix missed DMA unmap for aborted commands
  scsi: scsi_dh_alua: Properly handle the ALUA transitioning state

2 years agogfs2: Align read and write chunks to the page cache
Andreas Gruenbacher [Thu, 5 May 2022 11:32:23 +0000 (13:32 +0200)]
gfs2: Align read and write chunks to the page cache

Align the chunks that reads and writes are carried out in to the page
cache rather than the user buffers.  This will be more efficient in
general, especially for allocating writes.  Optimizing the case that the
user buffer is gfs2 backed isn't very useful; we only need to make sure
we won't deadlock.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agogfs2: Pull return value test out of should_fault_in_pages
Andreas Gruenbacher [Thu, 5 May 2022 10:53:26 +0000 (12:53 +0200)]
gfs2: Pull return value test out of should_fault_in_pages

Pull the return value test of the previous read or write operation out
of should_fault_in_pages().  In a following patch, we'll fault in pages
before the I/O and there will be no return value to check.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agogfs2: Clean up use of fault_in_iov_iter_{read,write}able
Andreas Gruenbacher [Thu, 5 May 2022 10:37:49 +0000 (12:37 +0200)]
gfs2: Clean up use of fault_in_iov_iter_{read,write}able

No need to store the return value of the fault_in functions in separate
variables.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agogfs2: Variable rename
Andreas Gruenbacher [Wed, 27 Apr 2022 11:53:42 +0000 (13:53 +0200)]
gfs2: Variable rename

Instead of counting the number of bytes read from the filesystem,
functions gfs2_file_direct_read and gfs2_file_read_iter count the number
of bytes written into the user buffer.  Conversely, functions
gfs2_file_direct_write and gfs2_file_buffered_write count the number of
bytes read from the user buffer.  This is nothing but confusing, so
change the read functions to count how many bytes they have read, and
the write functions to count how many bytes they have written.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agogfs2: Fix filesystem block deallocation for short writes
Andreas Gruenbacher [Thu, 14 Apr 2022 15:52:39 +0000 (17:52 +0200)]
gfs2: Fix filesystem block deallocation for short writes

When a write cannot be carried out in full, gfs2_iomap_end() releases
blocks that have been allocated for this write but haven't been used.

To compute the end of the allocation, gfs2_iomap_end() incorrectly
rounded the end of the attempted write down to the next block boundary
to arrive at the end of the allocation.  It would have to round up, but
the end of the allocation is also available as iomap->offset +
iomap->length, so just use that instead.

In addition, use round_up() for computing the start of the unused range.

Fixes: 64bc06bb32ee ("gfs2: iomap buffered write support")
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
2 years agoMerge tag 'ceph-for-5.18-rc7' of https://github.com/ceph/ceph-client
Linus Torvalds [Fri, 13 May 2022 18:12:04 +0000 (11:12 -0700)]
Merge tag 'ceph-for-5.18-rc7' of https://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "Two fixes to properly maintain xattrs on async creates and thus
  preserve SELinux context on newly created files and to avoid improper
  usage of folio->private field which triggered BUG_ONs.

  Both marked for stable"

* tag 'ceph-for-5.18-rc7' of https://github.com/ceph/ceph-client:
  ceph: check folio PG_private bit instead of folio->private
  ceph: fix setting of xattrs on async created inodes

2 years agoMerge tag 'nfs-for-5.18-4' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Fri, 13 May 2022 18:04:37 +0000 (11:04 -0700)]
Merge tag 'nfs-for-5.18-4' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:
 "One more pull request. There was a bug in the fix to ensure that gss-
  proxy continues to work correctly after we fixed the AF_LOCAL socket
  leak in the RPC code. This therefore reverts that broken patch, and
  replaces it with one that works correctly.

  Stable fixes:

   - SUNRPC: Ensure that the gssproxy client can start in a connected
     state

  Bugfixes:

   - Revert "SUNRPC: Ensure gss-proxy connects on setup"

   - nfs: fix broken handling of the softreval mount option"

* tag 'nfs-for-5.18-4' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  nfs: fix broken handling of the softreval mount option
  SUNRPC: Ensure that the gssproxy client can start in a connected state
  Revert "SUNRPC: Ensure gss-proxy connects on setup"

2 years agoMerge tag 'mm-hotfixes-stable-2022-05-11' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Fri, 13 May 2022 17:22:37 +0000 (10:22 -0700)]
Merge tag 'mm-hotfixes-stable-2022-05-11' of git://git./linux/kernel/git/akpm/mm

Pull misc fixes from Andrew Morton:
 "Seven MM fixes, three of which address issues added in the most recent
  merge window, four of which are cc:stable.

  Three non-MM fixes, none very serious"

[ And yes, that's a real pull request from Andrew, not me creating a
  branch from emailed patches. Woo-hoo! ]

* tag 'mm-hotfixes-stable-2022-05-11' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
  MAINTAINERS: add a mailing list for DAMON development
  selftests: vm: Makefile: rename TARGETS to VMTARGETS
  mm/kfence: reset PG_slab and memcg_data before freeing __kfence_pool
  mailmap: add entry for martyna.szapar-mudlaw@intel.com
  arm[64]/memremap: don't abuse pfn_valid() to ensure presence of linear map
  procfs: prevent unprivileged processes accessing fdinfo dir
  mm: mremap: fix sign for EFAULT error return value
  mm/hwpoison: use pr_err() instead of dump_page() in get_any_page()
  mm/huge_memory: do not overkill when splitting huge_zero_page
  Revert "mm/memory-failure.c: skip huge_zero_page in memory_failure()"

2 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 13 May 2022 17:17:39 +0000 (10:17 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:

 - TLB invalidation workaround for Qualcomm Kryo-4xx "gold" CPUs

 - Fix broken dependency in the vDSO Makefile

 - Fix pointer authentication overrides in ISAR2 ID register

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: Enable repeat tlbi workaround on KRYO4XX gold CPUs
  arm64: cpufeature: remove duplicate ID_AA64ISAR2_EL1 entry
  arm64: vdso: fix makefile dependency on vdso.so

2 years agoMerge tag 'hwmon-for-v5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 13 May 2022 17:10:07 +0000 (10:10 -0700)]
Merge tag 'hwmon-for-v5.18-rc7' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:

 - Restrict ltq-cputemp to SOC_XWAY to fix build failure

 - Add OF device ID table to tmp401 driver to enable auto-load

* tag 'hwmon-for-v5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (ltq-cputemp) restrict it to SOC_XWAY
  hwmon: (tmp401) Add OF device ID table

2 years agoMerge tag 'drm-fixes-2022-05-13' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 13 May 2022 17:00:37 +0000 (10:00 -0700)]
Merge tag 'drm-fixes-2022-05-13' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Pretty quiet week on the fixes front, 4 amdgpu and one i915 fix.

  I think there might be a few misc fbdev ones outstanding, but I'll see
  if they are necessary and pass them on if so.

  amdgpu:

   - Disable ASPM for VI boards on ADL platforms

   - S0ix DCN3.1 display fix

   - Resume regression fix

   - Stable pstate fix

  i915:

   - fix for kernel memory corruption when running a lot of OpenCL tests
     in parallel"

* tag 'drm-fixes-2022-05-13' of git://anongit.freedesktop.org/drm/drm:
  drm/amdgpu/ctx: only reset stable pstate if the user changed it (v2)
  Revert "drm/amd/pm: keep the BACO feature enabled for suspend"
  drm/i915: Fix race in __i915_vma_remove_closed
  drm/amd/display: undo clearing of z10 related function pointers
  drm/amdgpu: vi: disable ASPM on Intel Alder Lake based systems

2 years agodrm/vmwgfx: Disable command buffers on svga3 without gbobjects
Zack Rusin [Fri, 18 Mar 2022 17:43:31 +0000 (13:43 -0400)]
drm/vmwgfx: Disable command buffers on svga3 without gbobjects

With very limited vram on svga3 it's difficult to handle all the surface
migrations. Without gbobjects, i.e. the ability to store surfaces in
guest mobs, there's no reason to support intermediate svga2 features,
especially because we can fall back to fb traces and svga3 will never
support those in-between features.

On svga3 we wither want to use fb traces or screen targets
(i.e. gbobjects), nothing in between. This fixes presentation on a lot
of fusion/esxi tech previews where the exposed svga3 caps haven't been
finalized yet.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Fixes: 2cd80dbd3551 ("drm/vmwgfx: Add basic support for SVGA3")
Cc: <stable@vger.kernel.org> # v5.14+
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220318174332.440068-5-zack@kde.org
2 years agodrm/vmwgfx: Initialize drm_mode_fb_cmd2
Zack Rusin [Wed, 2 Mar 2022 15:24:24 +0000 (10:24 -0500)]
drm/vmwgfx: Initialize drm_mode_fb_cmd2

Transition to drm_mode_fb_cmd2 from drm_mode_fb_cmd left the structure
unitialized. drm_mode_fb_cmd2 adds a few additional members, e.g. flags
and modifiers which were never initialized. Garbage in those members
can cause random failures during the bringup of the fbcon.

Initializing the structure fixes random blank screens after bootup due
to flags/modifiers mismatches during the fbcon bring up.

Fixes: dabdcdc9822a ("drm/vmwgfx: Switch to mode_cmd2")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: <stable@vger.kernel.org> # v4.10+
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Reviewed-by: Maaz Mombasawala <mombasawalam@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220302152426.885214-7-zack@kde.org
2 years agodrm/vmwgfx: Fix fencing on SVGAv3
Zack Rusin [Wed, 2 Mar 2022 15:24:22 +0000 (10:24 -0500)]
drm/vmwgfx: Fix fencing on SVGAv3

Port of the vmwgfx to SVGAv3 lacked support for fencing. SVGAv3 removed
FIFO's and replaced them with command buffers and extra registers.
The initial version of SVGAv3 lacked support for most advanced features
(e.g. 3D) which made fences unnecessary. That is no longer the case,
especially as 3D support is being turned on.

Switch from FIFO commands and capabilities to command buffers and extra
registers to enable fences on SVGAv3.

Fixes: 2cd80dbd3551 ("drm/vmwgfx: Add basic support for SVGA3")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Reviewed-by: Maaz Mombasawala <mombasawalam@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220302152426.885214-5-zack@kde.org
2 years agoMerge tag 'icc-5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/djakov...
Greg Kroah-Hartman [Fri, 13 May 2022 14:15:28 +0000 (16:15 +0200)]
Merge tag 'icc-5.18-rc6' of git://git./linux/kernel/git/djakov/icc into char-misc-linus

Pull interconnect fixes from Georgi:
 "interconnect fixes for v5.18-rc

  This contains an additional fix for sc7180 and sdx55 platforms that helps
  them to enter suspend even on devices that don't have the most recent DT
  changes.

  - interconnect: Restore sync state by ignoring ipa-virt in provider count

Signed-off-by: Georgi Djakov <djakov@kernel.org>"
* tag 'icc-5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/djakov/icc:
  interconnect: Restore sync state by ignoring ipa-virt in provider count

2 years agox86/mm: Fix marking of unused sub-pmd ranges
Adrian-Ken Rueegsegger [Mon, 9 May 2022 09:06:37 +0000 (11:06 +0200)]
x86/mm: Fix marking of unused sub-pmd ranges

The unused part precedes the new range spanned by the start, end parameters
of vmemmap_use_new_sub_pmd(). This means it actually goes from
ALIGN_DOWN(start, PMD_SIZE) up to start.

Use the correct address when applying the mark using memset.

Fixes: 8d400913c231 ("x86/vmemmap: handle unpopulated sub-pmd ranges")
Signed-off-by: Adrian-Ken Rueegsegger <ken@codelabs.ch>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220509090637.24152-2-ken@codelabs.ch
2 years agoMerge tag 'usb-serial-5.18-rc7' of https://git.kernel.org/pub/scm/linux/kernel/git...
Greg Kroah-Hartman [Fri, 13 May 2022 06:29:40 +0000 (08:29 +0200)]
Merge tag 'usb-serial-5.18-rc7' of https://git./linux/kernel/git/johan/usb-serial

Johan writes:

USB-serial fixes for 5.18-rc7

Here are some new device ids.

All have been in linux-next with no reported issues.

* tag 'usb-serial-5.18-rc7' of https://git.kernel.org/pub/scm/linux/kernel/git/johan/usb-serial:
  USB: serial: qcserial: add support for Sierra Wireless EM7590
  USB: serial: option: add Fibocom MA510 modem
  USB: serial: option: add Fibocom L610 modem
  USB: serial: pl2303: add device id for HP LM930 Display

2 years agoMerge tag 'amd-drm-fixes-5.18-2022-05-11' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Fri, 13 May 2022 00:40:55 +0000 (10:40 +1000)]
Merge tag 'amd-drm-fixes-5.18-2022-05-11' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.18-2022-05-11:

amdgpu:
- Disable ASPM for VI boards on ADL platforms
- S0ix DCN3.1 display fix
- Resume regression fix
- Stable pstate fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220511174422.5769-1-alexander.deucher@amd.com
2 years agoMerge tag 'drm-intel-fixes-2022-05-12' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Thu, 12 May 2022 23:24:44 +0000 (09:24 +1000)]
Merge tag 'drm-intel-fixes-2022-05-12' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

Fix for #5732: (Cc stable) kernel memory corruption when running a lot of OpenCL tests in parallel

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/YnykW6L4e7vD3yl3@jlahtine-mobl.ger.corp.intel.com
2 years agoMerge tag 'net-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 12 May 2022 18:51:45 +0000 (11:51 -0700)]
Merge tag 'net-5.18-rc7' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from wireless, and bluetooth.

  No outstanding fires.

  Current release - regressions:

   - eth: atlantic: always deep reset on pm op, fix null-deref

  Current release - new code bugs:

   - rds: use maybe_get_net() when acquiring refcount on TCP sockets
     [refinement of a previous fix]

   - eth: ocelot: mark traps with a bool instead of guessing type based
     on list membership

  Previous releases - regressions:

   - net: fix skipping features in for_each_netdev_feature()

   - phy: micrel: fix null-derefs on suspend/resume and probe

   - bcmgenet: check for Wake-on-LAN interrupt probe deferral

  Previous releases - always broken:

   - ipv4: drop dst in multicast routing path, prevent leaks

   - ping: fix address binding wrt vrf

   - net: fix wrong network header length when BPF protocol translation
     is used on skbs with a fraglist

   - bluetooth: fix the creation of hdev->name

   - rfkill: uapi: fix RFKILL_IOCTL_MAX_SIZE ioctl request definition

   - wifi: iwlwifi: iwl-dbg: use del_timer_sync() before freeing

   - wifi: ath11k: reduce the wait time of 11d scan and hw scan while
     adding an interface

   - mac80211: fix rx reordering with non explicit / psmp ack policy

   - mac80211: reset MBSSID parameters upon connection

   - nl80211: fix races in nl80211_set_tx_bitrate_mask()

   - tls: fix context leak on tls_device_down

   - sched: act_pedit: really ensure the skb is writable

   - batman-adv: don't skb_split skbuffs with frag_list

   - eth: ocelot: fix various issues with TC actions (null-deref; bad
     stats; ineffective drops; ineffective filter removal)"

* tag 'net-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (61 commits)
  tls: Fix context leak on tls_device_down
  net: sfc: ef10: fix memory leak in efx_ef10_mtd_probe()
  net/smc: non blocking recvmsg() return -EAGAIN when no data and signal_pending
  net: dsa: bcm_sf2: Fix Wake-on-LAN with mac_link_down()
  mlxsw: Avoid warning during ip6gre device removal
  net: bcmgenet: Check for Wake-on-LAN interrupt probe deferral
  net: ethernet: mediatek: ppe: fix wrong size passed to memset()
  Bluetooth: Fix the creation of hdev->name
  i40e: i40e_main: fix a missing check on list iterator
  net/sched: act_pedit: really ensure the skb is writable
  s390/lcs: fix variable dereferenced before check
  s390/ctcm: fix potential memory leak
  s390/ctcm: fix variable dereferenced before check
  net: atlantic: verify hw_head_ lies within TX buffer ring
  net: atlantic: add check for MAX_SKB_FRAGS
  net: atlantic: reduce scope of is_rsc_complete
  net: atlantic: fix "frag[0] not initialized"
  net: stmmac: fix missing pci_disable_device() on error in stmmac_pci_probe()
  net: phy: micrel: Fix incorrect variable type in micrel
  decnet: Use container_of() for struct dn_neigh casts
  ...

2 years agoMerge branch 'for-5.18-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
Linus Torvalds [Thu, 12 May 2022 17:42:56 +0000 (10:42 -0700)]
Merge branch 'for-5.18-fixes' of git://git./linux/kernel/git/tj/cgroup

Pull cgroup fix from Tejun Heo:
 "Waiman's fix for a cgroup2 cpuset bug where it could miss nodes which
  were hot-added"

* 'for-5.18-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup/cpuset: Remove cpus_allowed/mems_allowed setup in cpuset_init_smp()

2 years agoMerge tag 'fixes_for_v5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 12 May 2022 17:21:44 +0000 (10:21 -0700)]
Merge tag 'fixes_for_v5.18-rc7' of git://git./linux/kernel/git/jack/linux-fs

Pull fs fixes from Jan Kara:
 "Three fixes that I'd still like to get to 5.18:

   - add a missing sanity check in the fanotify FAN_RENAME feature
     (added in 5.17, let's fix it before it gets wider usage in
     userspace)

   - udf fix for recently introduced filesystem corruption issue

   - writeback fix for a race in inode list handling that can lead to
     delayed writeback and possible dirty throttling stalls"

* tag 'fixes_for_v5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  udf: Avoid using stale lengthOfImpUse
  writeback: Avoid skipping inode writeback
  fanotify: do not allow setting dirent events in mask of non-dir

2 years agotls: Fix context leak on tls_device_down
Maxim Mikityanskiy [Thu, 12 May 2022 09:18:30 +0000 (12:18 +0300)]
tls: Fix context leak on tls_device_down

The commit cited below claims to fix a use-after-free condition after
tls_device_down. Apparently, the description wasn't fully accurate. The
context stayed alive, but ctx->netdev became NULL, and the offload was
torn down without a proper fallback, so a bug was present, but a
different kind of bug.

Due to misunderstanding of the issue, the original patch dropped the
refcount_dec_and_test line for the context to avoid the alleged
premature deallocation. That line has to be restored, because it matches
the refcount_inc_not_zero from the same function, otherwise the contexts
that survived tls_device_down are leaked.

This patch fixes the described issue by restoring refcount_dec_and_test.
After this change, there is no leak anymore, and the fallback to
software kTLS still works.

Fixes: c55dcdd435aa ("net/tls: Fix use-after-free after the TLS device goes down and up")
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20220512091830.678684-1-maximmi@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: sfc: ef10: fix memory leak in efx_ef10_mtd_probe()
Taehee Yoo [Thu, 12 May 2022 05:47:09 +0000 (05:47 +0000)]
net: sfc: ef10: fix memory leak in efx_ef10_mtd_probe()

In the NIC ->probe() callback, ->mtd_probe() callback is called.
If NIC has 2 ports, ->probe() is called twice and ->mtd_probe() too.
In the ->mtd_probe(), which is efx_ef10_mtd_probe() it allocates and
initializes mtd partiion.
But mtd partition for sfc is shared data.
So that allocated mtd partition data from last called
efx_ef10_mtd_probe() will not be used.
Therefore it must be freed.
But it doesn't free a not used mtd partition data in efx_ef10_mtd_probe().

kmemleak reports:
unreferenced object 0xffff88811ddb0000 (size 63168):
  comm "systemd-udevd", pid 265, jiffies 4294681048 (age 348.586s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffffa3767749>] kmalloc_order_trace+0x19/0x120
    [<ffffffffa3873f0e>] __kmalloc+0x20e/0x250
    [<ffffffffc041389f>] efx_ef10_mtd_probe+0x11f/0x270 [sfc]
    [<ffffffffc0484c8a>] efx_pci_probe.cold.17+0x3df/0x53d [sfc]
    [<ffffffffa414192c>] local_pci_probe+0xdc/0x170
    [<ffffffffa4145df5>] pci_device_probe+0x235/0x680
    [<ffffffffa443dd52>] really_probe+0x1c2/0x8f0
    [<ffffffffa443e72b>] __driver_probe_device+0x2ab/0x460
    [<ffffffffa443e92a>] driver_probe_device+0x4a/0x120
    [<ffffffffa443f2ae>] __driver_attach+0x16e/0x320
    [<ffffffffa4437a90>] bus_for_each_dev+0x110/0x190
    [<ffffffffa443b75e>] bus_add_driver+0x39e/0x560
    [<ffffffffa4440b1e>] driver_register+0x18e/0x310
    [<ffffffffc02e2055>] 0xffffffffc02e2055
    [<ffffffffa3001af3>] do_one_initcall+0xc3/0x450
    [<ffffffffa33ca574>] do_init_module+0x1b4/0x700

Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Fixes: 8127d661e77f ("sfc: Add support for Solarflare SFC9100 family")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Link: https://lore.kernel.org/r/20220512054709.12513-1-ap420073@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/smc: non blocking recvmsg() return -EAGAIN when no data and signal_pending
Guangguan Wang [Thu, 12 May 2022 03:08:20 +0000 (11:08 +0800)]
net/smc: non blocking recvmsg() return -EAGAIN when no data and signal_pending

Non blocking sendmsg will return -EAGAIN when any signal pending
and no send space left, while non blocking recvmsg return -EINTR
when signal pending and no data received. This may makes confused.
As TCP returns -EAGAIN in the conditions described above. Align the
behavior of smc with TCP.

Fixes: 846e344eb722 ("net/smc: add receive timeout check")
Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Link: https://lore.kernel.org/r/20220512030820.73848-1-guangguan.wang@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: bcm_sf2: Fix Wake-on-LAN with mac_link_down()
Florian Fainelli [Thu, 12 May 2022 02:17:31 +0000 (19:17 -0700)]
net: dsa: bcm_sf2: Fix Wake-on-LAN with mac_link_down()

After commit 2d1f90f9ba83 ("net: dsa/bcm_sf2: fix incorrect usage of
state->link") the interface suspend path would call our mac_link_down()
call back which would forcibly set the link down, thus preventing
Wake-on-LAN packets from reaching our management port.

Fix this by looking at whether the port is enabled for Wake-on-LAN and
not clearing the link status in that case to let packets go through.

Fixes: 2d1f90f9ba83 ("net: dsa/bcm_sf2: fix incorrect usage of state->link")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20220512021731.2494261-1-f.fainelli@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agousb: xhci-mtk: remove bandwidth budget table
Chunfeng Yun [Thu, 12 May 2022 06:49:31 +0000 (14:49 +0800)]
usb: xhci-mtk: remove bandwidth budget table

The bandwidth budget table is introduced to trace ideal bandwidth used
by each INT/ISOC endpoint, but in fact the endpoint may consume more
bandwidth and cause data transfer error, so it's better to leave some
margin. Obviously it's difficult to find the best margin for all cases,
instead take use of the worst-case scenario.

Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Link: https://lore.kernel.org/r/20220512064931.31670-2-chunfeng.yun@mediatek.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agousb: xhci-mtk: fix fs isoc's transfer error
Chunfeng Yun [Thu, 12 May 2022 06:49:30 +0000 (14:49 +0800)]
usb: xhci-mtk: fix fs isoc's transfer error

Due to the scheduler allocates the optimal bandwidth for FS ISOC endpoints,
this may be not enough actually and causes data transfer error, so come up
with an estimate that is no less than the worst case bandwidth used for
any one mframe, but may be an over-estimate.

Fixes: 451d3912586a ("usb: xhci-mtk: update fs bus bandwidth by bw_budget_table")
Cc: stable@vger.kernel.org
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Link: https://lore.kernel.org/r/20220512064931.31670-1-chunfeng.yun@mediatek.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agousb: gadget: fix race when gadget driver register via ioctl
Schspa Shi [Sun, 8 May 2022 15:02:47 +0000 (23:02 +0800)]
usb: gadget: fix race when gadget driver register via ioctl

The usb_gadget_register_driver can be called multi time by to
threads via USB_RAW_IOCTL_RUN ioctl syscall, which will lead
to multiple registrations.

Call trace:
  driver_register+0x220/0x3a0 drivers/base/driver.c:171
  usb_gadget_register_driver_owner+0xfb/0x1e0
    drivers/usb/gadget/udc/core.c:1546
  raw_ioctl_run drivers/usb/gadget/legacy/raw_gadget.c:513 [inline]
  raw_ioctl+0x1883/0x2730 drivers/usb/gadget/legacy/raw_gadget.c:1220
  ioctl USB_RAW_IOCTL_RUN

This routine allows two processes to register the same driver instance
via ioctl syscall. which lead to a race condition.

Please refer to the following scenarios.

           T1                                  T2
------------------------------------------------------------------
usb_gadget_register_driver_owner
  driver_register                    driver_register
    driver_find                       driver_find
    bus_add_driver                    bus_add_driver
      priv alloced                     <context switch>
      drv->p = priv;
      <schedule out>
      kobject_init_and_add // refcount = 1;
   //couldn't find an available UDC or it's busy
   <context switch>
                                       priv alloced
                                       drv->priv = priv;
                                       kobject_init_and_add
                                         ---> refcount = 1 <------
                                       // register success
                                       <context switch>
===================== another ioctl/process ======================
                                      driver_register
                                       driver_find
                                        k = kset_find_obj()
                                         ---> refcount = 2 <------
                                        <context out>
   driver_unregister
   // drv->p become T2's priv
   ---> refcount = 1 <------
   <context switch>
                                        kobject_put(k)
                                         ---> refcount = 0 <------
                                        return priv->driver;
                                        --------UAF here----------

There will be UAF in this scenario.

We can fix it by adding a new STATE_DEV_REGISTERING device state to
avoid double register.

Reported-by: syzbot+dc7c3ca638e773db07f6@syzkaller.appspotmail.com
Link: https://lore.kernel.org/all/000000000000e66c2805de55b15a@google.com/
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: Schspa Shi <schspa@gmail.com>
Link: https://lore.kernel.org/r/20220508150247.38204-1-schspa@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>