LoongArch: KVM: irqchip fixes
Subsystems (3)
v5 cover LoongArch: KVM: irqchip fixes
From: Tao Cui <[email protected]>
Hi,
Three fixes for the LoongArch KVM irqchip code:
- Patch 1 clears the device pointer in the destroy callbacks: when KVM_CREATE_DEVICE succeeds but the following fd allocation fails (e.g. under RLIMIT_NOFILE), ops->destroy() frees the irqchip while kvm->arch.* still points to it.
- Patch 2 aligns kvm_pch_pic_create() with kvm_eiointc_create() by propagating the real error code; the kvm_ipi_create() counterpart is being fixed separately (Chaithanya Lagisetty).
- Patch 3 rejects repeated PCH-PIC CTRL_INIT with -EEXIST, using the initialized state of the kvm_io_device as the indicator. The check, the MMIO base update and the bus registration are serialized by slots_lock; on a registration failure the state is rolled back so that a retry can start over.
All patches carry Fixes tags.
Changes in v5:
- Patch 3: roll back the base and the device state on registration failure, as suggested by Bibo.
Changes in v4:
- Patch 3: update pch_pic_base unconditionally, as suggested by Huacai.
Changes in v3:
- Drop "Load dmsintc pointer once in pch_msi_set_irq" (v2 patch 2), suggested by Bibo and Huacai.
- Patch 3 (v2 patch 4): reject repeated CTRL_INIT by checking the kvm_io_device state instead of a separate has_init flag, also suggested by Huacai; the check stays under slots_lock.
Changes in v2:
- Drop "Guard against NULL irqchip in irq injection" (v1 patch 2) and "Rebase steal time counter in vcpu context" (v1 patch 4) after review discussion: their trigger scenarios are not reachable through real VMM behaviour.
- Use the one-line destroy style in dmsintc suggested by Bibo.
- Track repeated PCH-PIC CTRL_INIT with a has_init flag and return -EEXIST instead of -EBUSY, also suggested by Bibo; move the check and base update under slots_lock so they are atomic.
- Add READ_ONCE() to the dmsintc pointer loads so the compiler keeps each of them a single load.
Tao Cui (3):
LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
LoongArch: KVM: Propagate real error code in kvm_pch_pic_create
LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
arch/loongarch/kvm/intc/dmsintc.c | 1 + arch/loongarch/kvm/intc/eiointc.c | 1 + arch/loongarch/kvm/intc/ipi.c | 1 + arch/loongarch/kvm/intc/pch_pic.c | 19 +++++++++++++++---- 4 files changed, 18 insertions(+), 4 deletions(-)
Signature · 1 line
2.43.0
PATCH v5 1/3 LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
From: Tao Cui <[email protected]>
The destroy callbacks of the four irqchip devices free the device structure without clearing kvm->arch.{ipi,eiointc,pch_pic,dmsintc}, leaving a dangling pointer. ops->destroy() is not only called at VM teardown but also when KVM_CREATE_DEVICE succeeds and the following anon_inode_getfd() fails (e.g. under RLIMIT_NOFILE); the VM stays alive, kvm_arch_irqchip_in_kernel() still reports true, and interrupt injection dereferences the freed device.
Clear the pointer when destroying.
--- arch/loongarch/kvm/intc/dmsintc.c | 1 + arch/loongarch/kvm/intc/eiointc.c | 1 + arch/loongarch/kvm/intc/ipi.c | 1 + arch/loongarch/kvm/intc/pch_pic.c | 1 + 4 files changed, 4 insertions(+)
arch/loongarch/kvm/intc/dmsintc.c +1 -0
@@ -146,6 +146,7 @@ static void kvm_dmsintc_destroy(struct kvm_device *dev)return ;kfree (dev->kvm->arch.dmsintc); + dev->kvm->arch.dmsintc =NULL ;kfree (dev); }
arch/loongarch/kvm/intc/eiointc.c +1 -0
@@ -675,6 +675,7 @@ static void kvm_eiointc_destroy(struct kvm_device *dev) kvm = dev->kvm; eiointc = kvm->arch.eiointc; + kvm->arch.eiointc =NULL ;mutex_lock (&kvm->slots_lock);kvm_io_bus_unregister_dev (kvm,KVM_IOCSR_BUS , &eiointc->device);kvm_io_bus_unregister_dev (kvm,KVM_IOCSR_BUS , &eiointc->device_vext);
arch/loongarch/kvm/intc/ipi.c +1 -0
@@ -441,6 +441,7 @@ static void kvm_ipi_destroy(struct kvm_device *dev) kvm = dev->kvm; ipi = kvm->arch.ipi; + kvm->arch.ipi =NULL ;mutex_lock (&kvm->slots_lock);kvm_io_bus_unregister_dev (kvm,KVM_IOCSR_BUS , &ipi->device);mutex_unlock (&kvm->slots_lock);
arch/loongarch/kvm/intc/pch_pic.c +1 -0
@@ -483,6 +483,7 @@ static void kvm_pch_pic_destroy(struct kvm_device *dev) kvm = dev->kvm; s = kvm->arch.pch_pic; + kvm->arch.pch_pic =NULL ;/* unregister pch pic device and free it's memory */ mutex_lock (&kvm->slots_lock);kvm_io_bus_unregister_dev (kvm,KVM_MMIO_BUS , &s->device);
Signature · 1 line
2.43.0
[email protected]
AI reviewRe: [PATCH v5 1/3] LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider: - [Critical] kvm: NULL pointer dereference in lockless IRQ injection paths - [High] kvm: loongarch: UAF…
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Critical] kvm: NULL pointer dereference in lockless IRQ injection paths
- [High] kvm: loongarch: UAF window in kvm_dmsintc_destroy due to wrong kfree order --
commit b63c6600199f59512e0a9603666fb0590806637a
Author: Tao Cui <[email protected]>
LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
This commit clears the irqchip device pointers when destroying devices to prevent dangling pointers from being dereferenced if device creation fails later.
arch/loongarch/kvm/intc/dmsintc.c static void kvm_dmsintc_destroy(struct kvm_device *dev) · line 149
Quoted 9 lines
arch/loongarch/kvm/intc/dmsintc.c +1
-0@@ -146,6 +146,7 @@ static void kvm_dmsintc_destroy(struct kvm_device *dev)return ;kfree (dev->kvm->arch.dmsintc); + dev->kvm->arch.dmsintc =NULL ;
[Severity: High]
Does this create a Use-After-Free window? The memory is freed before the globally visible pointer is cleared. If KVM_CREATE_DEVICE succeeds but a subsequent error (e.g. in anon_inode_getfd) triggers ops->destroy(), this invokes kvm_dmsintc_destroy() while concurrent threads can still access the pointer.
Should the pointer be cleared before calling kfree(), as is correctly done for the other irqchip devices in this patch?
Quoted 3 lines
arch/loongarch/kvm/intc/dmsintc.c +0
-0kfree (dev); }
[ ... ]
arch/loongarch/kvm/intc/pch_pic.c static void kvm_pch_pic_destroy(struct kvm_device *dev) · line 486
Quoted 9 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0@@ -483,6 +483,7 @@ static void kvm_pch_pic_destroy(struct kvm_device *dev) kvm = dev->kvm; s = kvm->arch.pch_pic; + kvm->arch.pch_pic =NULL ;
[Severity: Critical]
Does setting this pointer to NULL introduce a NULL pointer dereference in the concurrent lockless IRQ injection paths?
If a user-space process maps GSI routing to an irqchip and races KVM_IRQ_LINE or KVM_IRQFD interrupt injection against a failing KVM_CREATE_DEVICE ioctl (e.g. by hitting the process RLIMIT_NOFILE limit), this teardown path will assign NULL to the irqchip pointer.
Concurrently, the IRQ injection path locklessly invokes routing callbacks like kvm_set_pic_irq() which pass the potentially NULL kvm->arch.pch_pic into pch_pic_set_irq():
arch/loongarch/kvm/irqfd.c:kvm_set_pic_irq() {
...
pch_pic_set_irq(kvm->arch.pch_pic, e->irqchip.pin, level);
...
}The pointer is then unconditionally dereferenced:
arch/loongarch/kvm/intc/pch_pic.c:pch_pic_set_irq() {
...
spin_lock(&s->lock);
...
}A similar pattern exists for eiointc_set_irq() inarch/loongarch/kvm/intc/eiointc.c.
Are NULL guard checks needed in the concurrent readers to safely handle this regression?
Quoted 3 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0/* unregister pch pic device and free it's memory */ mutex_lock (&kvm->slots_lock);kvm_io_bus_unregister_dev (kvm,KVM_MMIO_BUS , &s->device);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
PATCH v5 2/3 LoongArch: KVM: Propagate real error code in kvm_pch_pic_create
In reply to Tao CuiFrom: Tao Cui <[email protected]>
kvm_pch_pic_create() replaces the real error of kvm_setup_default_irq_routing() with a fixed -ENOMEM. Return the actual code, as kvm_eiointc_create() already does; the kvm_ipi_create() case is fixed separately by "Return the actual error code in kvm_ipi_create()" (Chaithanya Lagisetty).
--- arch/loongarch/kvm/intc/pch_pic.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/loongarch/kvm/intc/pch_pic.c +1 -1
@@ -444,7 +444,7 @@ static int kvm_pch_pic_create(struct kvm_device *dev, u32 type) ret =kvm_setup_default_irq_routing (kvm);if (ret) -return -ENOMEM ; +return ret; s = kzalloc_obj(struct loongarch_pch_pic);if (!s)
Signature · 1 line
2.43.0
PATCH v5 3/3 LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
In reply to Tao CuiFrom: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIT has no guard against repeated
invocation: every call overwrites pch_pic_base and registers the same kvm_io_device on the MMIO bus at the new address, while kvm_pch_pic_destroy() unregisters only one bus range. After a repeated init, MMIO to the stale ranges computes its register offset against the new base and silently reads 0 / drops writes, and the leftover bus entries persist until the VM is destroyed.
Reject repeated initialization with -EEXIST, using the initialized state of the kvm_io_device as the indicator. The check, the MMIO base update and the bus registration are serialized by slots_lock, and the real registration error is propagated instead of being replaced with -EFAULT. On a registration failure the base and the device state are rolled back so that a retry can start over.
--- arch/loongarch/kvm/intc/pch_pic.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-)
arch/loongarch/kvm/intc/pch_pic.c +13 -3
@@ -281,16 +281,26 @@ static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr)struct kvm_io_device *device;struct loongarch_pch_pic *s = dev->kvm->arch.pch_pic; - s->pch_pic_base = addr; device = &s->device; +mutex_lock (&kvm->slots_lock); +if (s->device.ops) { + ret = -EEXIST ; + goto out; + } + s->pch_pic_base = addr;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops); -mutex_lock (&kvm->slots_lock);/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (ret) { +/* undo the state so that a retry can start over */ + s->pch_pic_base =0 ; + s->device.ops =NULL ; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */
Signature · 1 line
2.43.0
Huacai Chen
maintainerSeries applied, thanks.
Huacai
Quoted 73 lines · Tao Cui
From: Tao Cui <[email protected]>
Hi,
Three fixes for the LoongArch KVM irqchip code:
- Patch 1 clears the device pointer in the destroy callbacks: when
KVM_CREATE_DEVICEsucceeds but the following fd allocation fails (e.g. underRLIMIT_NOFILE),ops->destroy()frees the irqchip while kvm->arch.* still points to it.- Patch 2 aligns
kvm_pch_pic_create()withkvm_eiointc_create()by propagating the real error code; thekvm_ipi_create()counterpart is being fixed separately (Chaithanya Lagisetty).- Patch 3 rejects repeated PCH-PIC
CTRL_INITwith-EEXIST, using the initialized state of thekvm_io_deviceas the indicator. The check, the MMIO base update and the bus registration are serialized byslots_lock; on a registration failure the state is rolled back so that a retry can start over.All patches carry Fixes tags.
Changes in v5:
- Patch 3: roll back the base and the device state on registration failure, as suggested by Bibo.
Changes in v4:
- Patch 3: update
pch_pic_baseunconditionally, as suggested by Huacai.Changes in v3:
- Drop "Load dmsintc pointer once in
pch_msi_set_irq" (v2 patch 2), suggested by Bibo and Huacai.- Patch 3 (v2 patch 4): reject repeated
CTRL_INITby checking thekvm_io_devicestate instead of a separatehas_initflag, also suggested by Huacai; the check stays underslots_lock.Changes in v2:
- Drop "Guard against NULL irqchip in irq injection" (v1 patch 2) and "Rebase steal time counter in vcpu context" (v1 patch 4) after review discussion: their trigger scenarios are not reachable through real VMM behaviour.
- Use the one-line destroy style in dmsintc suggested by Bibo.
- Track repeated PCH-PIC
CTRL_INITwith ahas_initflag and return-EEXISTinstead of-EBUSY, also suggested by Bibo; move the check and base update underslots_lockso they are atomic.- Add
READ_ONCE()to the dmsintc pointer loads so the compiler keeps each of them a single load.Tao Cui (3):
LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
LoongArch: KVM: Propagate real error code inkvm_pch_pic_create
LoongArch: KVM: Reject repeated PCH-PICCTRL_INITarch/loongarch/kvm/intc/dmsintc.c | 1 + arch/loongarch/kvm/intc/eiointc.c | 1 + arch/loongarch/kvm/intc/ipi.c | 1 + arch/loongarch/kvm/intc/pch_pic.c | 19 +++++++++++++++---- 4 files changed, 18 insertions(+), 4 deletions(-)--
2.43.0