LoongArch: KVM: irqchip fixes
Subsystems (3)
v2 cover LoongArch: KVM: irqchip fixes
From: Tao Cui <[email protected]>
Hi,
Four 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 loads kvm->arch.dmsintc once in the MSI injection path:
pch_msi_set_irq() re-reads the pointer between the non-NULL check and the address-window comparison, so a concurrent device removal can be observed between them.
- Patch 3 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 4 rejects repeated PCH-PIC CTRL_INIT with -EEXIST, tracking the state with a has_init flag so the check and the MMIO base update are atomic under slots_lock.
All patches carry Fixes tags.
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 (4):
LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
LoongArch: KVM: Load dmsintc pointer once in pch_msi_set_irq
LoongArch: KVM: Propagate real error code in kvm_pch_pic_create
LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/dmsintc.c | 7 ++++++- arch/loongarch/kvm/intc/eiointc.c | 1 + arch/loongarch/kvm/intc/ipi.c | 1 + arch/loongarch/kvm/intc/pch_pic.c | 22 ++++++++++++++++------ 5 files changed, 25 insertions(+), 7 deletions(-)
Signature · 1 line
2.43.0
PATCH v2 1/4 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
@@ -166,6 +166,7 @@ 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
@@ -444,6 +444,7 @@ 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
PATCH v2 2/4 LoongArch: KVM: Load dmsintc pointer once in pch_msi_set_irq
In reply to Tao CuiFrom: Tao Cui <[email protected]>
pch_msi_set_irq() reads kvm->arch.dmsintc several times: the non-NULL check and the address-window comparison each reload the pointer, so a concurrent device removal can be observed between them and the following dereference hits a freed object.
Load the pointer once at the top of pch_msi_set_irq() and add a NULL check with a local snapshot in dmsintc_set_irq(). Both loads use READ_ONCE() so the compiler keeps them single. This closes the reload race; a narrower window where removal happens right after the load remains, as the injection path takes no lock against destroy.
--- arch/loongarch/kvm/intc/dmsintc.c | 6 +++++- arch/loongarch/kvm/intc/pch_pic.c | 7 ++++--- 2 files changed, 9 insertions(+), 4 deletions(-)
arch/loongarch/kvm/intc/dmsintc.c +5 -1
@@ -69,9 +69,13 @@ int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) {unsigned int irq, cpu;struct kvm_vcpu *vcpu; +struct loongarch_dmsintc *s =READ_ONCE (kvm->arch.dmsintc); + +if (!s) +return -EINVAL ; irq = (addr >>AVEC_IRQ_SHIFT ) &AVEC_IRQ_MASK ; - cpu = (addr >>AVEC_CPU_SHIFT ) & kvm->arch.dmsintc->cpu_mask; + cpu = (addr >>AVEC_CPU_SHIFT ) & s->cpu_mask;if (cpu >=KVM_MAX_VCPUS )return -EINVAL ; vcpu =kvm_get_vcpu_by_cpuid (kvm, cpu);
arch/loongarch/kvm/intc/pch_pic.c +4 -3
@@ -71,10 +71,11 @@ void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level)int pch_msi_set_irq (struct kvm *kvm,struct kvm_kernel_irq_routing_entry *e,int level) {u64 msg_addr = (((u64 )e->msi.address_hi) <<32 ) | e->msi.address_lo; +struct loongarch_dmsintc *dmsintc =READ_ONCE (kvm->arch.dmsintc); -if (cpu_has_msgint && kvm->arch.dmsintc && - msg_addr >= kvm->arch.dmsintc->msg_addr_base && - msg_addr < (kvm->arch.dmsintc->msg_addr_base + kvm->arch.dmsintc->msg_addr_size)) { +if (cpu_has_msgint && dmsintc && + msg_addr >= dmsintc->msg_addr_base && + msg_addr < (dmsintc->msg_addr_base + dmsintc->msg_addr_size)) {return dmsintc_set_irq (kvm, msg_addr, e->msi.data, level); }
Signature · 1 line
2.43.0
PATCH v2 3/4 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
@@ -445,7 +445,7 @@ 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 v2 4/4 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, tracking the state with a has_init flag so the check and the MMIO base update are atomic under slots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with -EFAULT.
--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)
arch/loongarch/include/asm/kvm_pch_pic.h +1 -0
@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */
arch/loongarch/kvm/intc/pch_pic.c +10 -2
@@ -282,16 +282,24 @@ 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;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +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
[email protected]
AI reviewRe: [PATCH v2 1/4] LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
In reply to Tao CuiThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Critical] kvm: loongarch: NULL deref and UAF via lockless irqchip destruction --
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Critical] kvm: loongarch: NULL deref and UAF via lockless irqchip destruction --
commit 1d2bf4dc9e1262efde3462bede6b473d0a4a01e5
Author: Tao Cui <[email protected]>
LoongArch: KVM: Clear device pointer in irqchip destroy callbacks
The commit clears device pointers in the destroy callbacks for the four LoongArch irqchip devices to prevent dangling pointers. This addresses an issue where anon_inode_getfd() failures during KVM_CREATE_DEVICE can leave the VM alive with a freed irqchip device.
arch/loongarch/kvm/intc/dmsintc.c static void kvm_dmsintc_destroy(struct kvm_device *dev) · line 169
Quoted 9 lines
arch/loongarch/kvm/intc/dmsintc.c +1
-0@@ -166,6 +166,7 @@ static void kvm_dmsintc_destroy(struct kvm_device *dev)return ;kfree (dev->kvm->arch.dmsintc); + dev->kvm->arch.dmsintc =NULL ;
[Severity: Critical]
Does setting this pointer to NULL after freeing it leave a window open for a use-after-free?
If anon_inode_getfd() fails during KVM_CREATE_DEVICE, the error path calls ops->destroy() locklessly. At the same time, concurrent KVM_IRQ_LINE requests can locklessly invoke kvm_set_irq(), which drops kvm->irq_srcu before calling the routing callbacks.
If those callbacks read the pointer between the kfree() and this NULL assignment, won't they dereference freed memory? Does this ordering defeat any READ_ONCE() fixes intended for concurrent lockless readers?
[ ... ]
arch/loongarch/kvm/intc/pch_pic.c static void kvm_pch_pic_destroy(struct kvm_device *dev) · line 483
Quoted 9 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0@@ -480,6 +480,7 @@ 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]
Can assigning NULL here without synchronization lead to a deterministic NULL pointer dereference and host panic?
If kvm->arch.pch_pic is set to NULL locklessly here, and a concurrent KVM_IRQ_LINE ioctl calls into kvm_set_pic_irq(), the routing table execution happens after kvm_set_irq() drops irq_srcu:
virt/kvm/irqchip.c:kvm_set_irq() {
...
idx = srcu_read_lock(&kvm->irq_srcu);
i = kvm_irq_map_gsi(kvm, irq_set, irq);
srcu_read_unlock(&kvm->irq_srcu, idx);
while (i--) {
int r;
r = irq_set[i].set(&irq_set[i], kvm, irq_source_id, level,
line_status);
...
}The callback kvm_set_pic_irq() then blindly passes the newly-NULL pointer to pch_pic_set_irq() without checking it:
arch/loongarch/kvm/irqfd.c:kvm_set_pic_irq() {
...
pch_pic_set_irq(kvm->arch.pch_pic, e->irqchip.pin, level);
...
}Since pch_pic_set_irq() unconditionally dereferences the pointer:
arch/loongarch/kvm/intc/pch_pic.c:pch_pic_set_irq() {
...
spin_lock(&s->lock);
...
}Doesn't this mean an unprivileged userspace process exhausting file descriptors and concurrently injecting interrupts will trigger a host panic?
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Huacai Chen
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
In reply to Tao CuiHi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;
Why so complicated? The below is enough, no?
(quoted diff) +0 -0
b/arch/loongarch/kvm/intc/pch_pic.c
index 2b63b0c2c7ce..7855d78304b7 100644
arch/loongarch/kvm/intc/pch_pic.c +0 -0
@@ -281,6 +281,9 @@ 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;+ if (s->device->ops)
+ return -EEXIST;
+
s->pch_pic_base = addr;
device = &s->device;
/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Re: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
Hi, Huacai.
Quoted 39 lines · Huacai Chen
Hi, Tao,
Quoted 36 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */
Quoted 9 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 7 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
Thanks for the suggestion.
I agree that a separate has_init flag can be unnecessary here. v3 uses s->device.ops to track whether the device has already been initialized.
The other changes are kept: pch_pic_base is updated only after successful bus registration, and the original registration error is returned instead of -EFAULT.
Thanks,
Tao
Quoted 44 lines
arch/loongarch/kvm/intc/pch_pic.c +3
-0@@ -281,6 +281,9 @@ 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; +if (s->device->ops) +return -EEXIST ; + s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Bibo Mao
maintainerRe: [PATCH v2 2/4] LoongArch: KVM: Load dmsintc pointer once in pch_msi_set_irq
In reply to Tao CuiQuoted 25 lines · Tao Cui
From: Tao Cui <[email protected]>
pch_msi_set_irq()readskvm->arch.dmsintcseveral times: the non-NULL check and the address-window comparison each reload the pointer, so a concurrent device removal can be observed between them and the following dereference hits a freed object.Load the pointer once at the top of
pch_msi_set_irq()and add a NULL check with a local snapshot indmsintc_set_irq(). Both loads useREAD_ONCE()so the compiler keeps them single. This closes the reload race; a narrower window where removal happens right after the load remains, as the injection path takes no lock against destroy.Fixes 03de5eecb0f0 ("LoongArch: KVM: Add DMSINTC inject msi to vCPU")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/kvm/intc/dmsintc.c | 6 +++++- arch/loongarch/kvm/intc/pch_pic.c | 7 ++++--- 2 files changed, 9 insertions(+), 4 deletions(-)arch/loongarch/kvm/intc/dmsintc.c +0
-0
arch/loongarch/kvm/intc/dmsintc.c int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) · line 75
Quoted 7 lines
arch/loongarch/kvm/intc/dmsintc.c +4
-0@@ -69,9 +69,13 @@ int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) {unsigned int irq, cpu;struct kvm_vcpu *vcpu; +struct loongarch_dmsintc *s =READ_ONCE (kvm->arch.dmsintc); + +if (!s) +return -EINVAL ;
NULL check with kvm->arch.dmsintc is already done in its caller function pch_msi_set_irq(). It is not necessary here.
Quoted 12 lines
arch/loongarch/kvm/intc/dmsintc.c +1
-1irq = (addr >>AVEC_IRQ_SHIFT ) &AVEC_IRQ_MASK ; - cpu = (addr >>AVEC_CPU_SHIFT ) & kvm->arch.dmsintc->cpu_mask; + cpu = (addr >>AVEC_CPU_SHIFT ) & s->cpu_mask;if (cpu >=KVM_MAX_VCPUS )return -EINVAL ; vcpu =kvm_get_vcpu_by_cpuid (kvm, cpu);arch/loongarch/kvm/intc/pch_pic.c +0
-0
arch/loongarch/kvm/intc/pch_pic.c void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level) · line 74
Quoted 4 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0@@ -71,10 +71,11 @@ void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level)int pch_msi_set_irq (struct kvm *kvm,struct kvm_kernel_irq_routing_entry *e,int level) {u64 msg_addr = (((u64 )e->msi.address_hi) <<32 ) | e->msi.address_lo; +struct loongarch_dmsintc *dmsintc =READ_ONCE (kvm->arch.dmsintc);
what is usage of READ_ONCE() here?
If you want to simple the usage of kvm->arch.dmsintc in multiple places, just *struct loongarch_dmsintc *dmsintc = kvm->arch.dmsintc* is enough. And it is not fixup patch, it is code cleanup.
Regards
Bibo Mao
Quoted 11 lines
arch/loongarch/kvm/intc/pch_pic.c +3
-3-if (cpu_has_msgint && kvm->arch.dmsintc && - msg_addr >= kvm->arch.dmsintc->msg_addr_base && - msg_addr < (kvm->arch.dmsintc->msg_addr_base + kvm->arch.dmsintc->msg_addr_size)) { +if (cpu_has_msgint && dmsintc && + msg_addr >= dmsintc->msg_addr_base && + msg_addr < (dmsintc->msg_addr_base + dmsintc->msg_addr_size)) {return dmsintc_set_irq (kvm, msg_addr, e->msi.data, level); }
Bibo Mao
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
In reply to Huacai ChenQuoted 54 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 6 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0@@ -281,6 +281,9 @@ 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; +if (s->device->ops) +return -EEXIST ;
This can work, however I think that it is not a good idea to access internal structure field about kvm_io_device. If so, there is no use about API kvm_iodevice_init(), just s->device->ops = &kvm_pch_pic_ops is ok.
If adding has_init is redundant, maybe we can set s->pch_pic_base with INVALID_GPA in kvm_pch_pic_create() or some other methods. However I think directly accessing kvm_io_device::ops is not a good method, no other architectures do in such way.
Regards
Bibo Mao
Quoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Huacai Chen
maintainerRe: [PATCH v2 2/4] LoongArch: KVM: Load dmsintc pointer once in pch_msi_set_irq
In reply to Bibo MaoQuoted 56 lines · Bibo Mao
Quoted 25 lines · Tao Cui
From: Tao Cui <[email protected]>
pch_msi_set_irq()readskvm->arch.dmsintcseveral times: the non-NULL check and the address-window comparison each reload the pointer, so a concurrent device removal can be observed between them and the following dereference hits a freed object.Load the pointer once at the top of
pch_msi_set_irq()and add a NULL check with a local snapshot indmsintc_set_irq(). Both loads useREAD_ONCE()so the compiler keeps them single. This closes the reload race; a narrower window where removal happens right after the load remains, as the injection path takes no lock against destroy.Fixes 03de5eecb0f0 ("LoongArch: KVM: Add DMSINTC inject msi to vCPU")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/kvm/intc/dmsintc.c | 6 +++++- arch/loongarch/kvm/intc/pch_pic.c | 7 ++++--- 2 files changed, 9 insertions(+), 4 deletions(-)arch/loongarch/kvm/intc/dmsintc.c +0
-0arch/loongarch/kvm/intc/dmsintc.c int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) · line 75
Quoted 7 lines
arch/loongarch/kvm/intc/dmsintc.c +4
-0@@ -69,9 +69,13 @@ int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) {unsigned int irq, cpu;struct kvm_vcpu *vcpu; +struct loongarch_dmsintc *s =READ_ONCE (kvm->arch.dmsintc); + +if (!s) +return -EINVAL ;NULL check with
kvm->arch.dmsintcis already done in its caller functionpch_msi_set_irq(). It is not necessary here.Quoted 12 lines
arch/loongarch/kvm/intc/dmsintc.c +1
-1irq = (addr >>AVEC_IRQ_SHIFT ) &AVEC_IRQ_MASK ; - cpu = (addr >>AVEC_CPU_SHIFT ) & kvm->arch.dmsintc->cpu_mask; + cpu = (addr >>AVEC_CPU_SHIFT ) & s->cpu_mask;if (cpu >=KVM_MAX_VCPUS )return -EINVAL ; vcpu =kvm_get_vcpu_by_cpuid (kvm, cpu);arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level) · line 74
Quoted 4 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0@@ -71,10 +71,11 @@ void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level)int pch_msi_set_irq (struct kvm *kvm,struct kvm_kernel_irq_routing_entry *e,int level) {u64 msg_addr = (((u64 )e->msi.address_hi) <<32 ) | e->msi.address_lo; +struct loongarch_dmsintc *dmsintc =READ_ONCE (kvm->arch.dmsintc);what is usage of
READ_ONCE()here?
Quoted 3 lines
If you want to simple the usage of
kvm->arch.dmsintcin multiple places, just *structloongarch_dmsintc*dmsintc = kvm->arch.dmsintc* is enough. And it is not fixup patch, it is code cleanup.
I completely don't think this patch is necessary.
Huacai
Quoted 15 lines
Regards
Bibo MaoQuoted 11 lines
arch/loongarch/kvm/intc/pch_pic.c +3
-3-if (cpu_has_msgint && kvm->arch.dmsintc && - msg_addr >= kvm->arch.dmsintc->msg_addr_base && - msg_addr < (kvm->arch.dmsintc->msg_addr_base + kvm->arch.dmsintc->msg_addr_size)) { +if (cpu_has_msgint && dmsintc && + msg_addr >= dmsintc->msg_addr_base && + msg_addr < (dmsintc->msg_addr_base + dmsintc->msg_addr_size)) {return dmsintc_set_irq (kvm, msg_addr, e->msi.data, level); }
Bibo Mao
maintainerRe: [PATCH v2 2/4] LoongArch: KVM: Load dmsintc pointer once in pch_msi_set_irq
Quoted 57 lines · Huacai Chen
Quoted 56 lines · Bibo Mao
Quoted 25 lines · Tao Cui
From: Tao Cui <[email protected]>
pch_msi_set_irq()readskvm->arch.dmsintcseveral times: the non-NULL check and the address-window comparison each reload the pointer, so a concurrent device removal can be observed between them and the following dereference hits a freed object.Load the pointer once at the top of
pch_msi_set_irq()and add a NULL check with a local snapshot indmsintc_set_irq(). Both loads useREAD_ONCE()so the compiler keeps them single. This closes the reload race; a narrower window where removal happens right after the load remains, as the injection path takes no lock against destroy.Fixes 03de5eecb0f0 ("LoongArch: KVM: Add DMSINTC inject msi to vCPU")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/kvm/intc/dmsintc.c | 6 +++++- arch/loongarch/kvm/intc/pch_pic.c | 7 ++++--- 2 files changed, 9 insertions(+), 4 deletions(-)arch/loongarch/kvm/intc/dmsintc.c +0
-0arch/loongarch/kvm/intc/dmsintc.c int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) · line 75
Quoted 7 lines
arch/loongarch/kvm/intc/dmsintc.c +4
-0@@ -69,9 +69,13 @@ int dmsintc_set_irq(struct kvm *kvm, u64 addr, int data, int level) {unsigned int irq, cpu;struct kvm_vcpu *vcpu; +struct loongarch_dmsintc *s =READ_ONCE (kvm->arch.dmsintc); + +if (!s) +return -EINVAL ;NULL check with
kvm->arch.dmsintcis already done in its caller functionpch_msi_set_irq(). It is not necessary here.Quoted 12 lines
arch/loongarch/kvm/intc/dmsintc.c +1
-1irq = (addr >>AVEC_IRQ_SHIFT ) &AVEC_IRQ_MASK ; - cpu = (addr >>AVEC_CPU_SHIFT ) & kvm->arch.dmsintc->cpu_mask; + cpu = (addr >>AVEC_CPU_SHIFT ) & s->cpu_mask;if (cpu >=KVM_MAX_VCPUS )return -EINVAL ; vcpu =kvm_get_vcpu_by_cpuid (kvm, cpu);arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level) · line 74
Quoted 4 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0@@ -71,10 +71,11 @@ void pch_pic_set_irq(struct loongarch_pch_pic *s, int irq, int level)int pch_msi_set_irq (struct kvm *kvm,struct kvm_kernel_irq_routing_entry *e,int level) {u64 msg_addr = (((u64 )e->msi.address_hi) <<32 ) | e->msi.address_lo; +struct loongarch_dmsintc *dmsintc =READ_ONCE (kvm->arch.dmsintc);what is usage of
READ_ONCE()here?
Quoted 4 lines
Quoted 3 lines
If you want to simple the usage of
kvm->arch.dmsintcin multiple places, just *structloongarch_dmsintc*dmsintc = kvm->arch.dmsintc* is enough. And it is not fixup patch, it is code cleanup.I completely don't think this patch is necessary.
yeap, I have the same feeling about this :)
Quoted 18 lines
Huacai
Quoted 15 lines
Regards
Bibo MaoQuoted 11 lines
arch/loongarch/kvm/intc/pch_pic.c +3
-3-if (cpu_has_msgint && kvm->arch.dmsintc && - msg_addr >= kvm->arch.dmsintc->msg_addr_base && - msg_addr < (kvm->arch.dmsintc->msg_addr_base + kvm->arch.dmsintc->msg_addr_size)) { +if (cpu_has_msgint && dmsintc && + msg_addr >= dmsintc->msg_addr_base && + msg_addr < (dmsintc->msg_addr_base + dmsintc->msg_addr_size)) {return dmsintc_set_irq (kvm, msg_addr, e->msi.data, level); }
Huacai Chen
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
In reply to Bibo MaoQuoted 62 lines · Bibo Mao
Quoted 58 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -281,6 +281,9 @@ 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;
Quoted 5 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0+if (s->device->ops) +return -EEXIST ;This can work, however I think that it is not a good idea to access internal structure field about
kvm_io_device. If so, there is no use about APIkvm_iodevice_init(), justs->device->ops=&kvm_pch_pic_opsis ok.
I'm a little not agree. :)
I think kvm_iodevice_init() is designed to do more work rather than just set the ops (though it just set the ops now), otherwise its name should be kvm_iodevice_set_ops().
In addition, even if kvm_iodevice_init() is really a setter, there is no getter for the ops, so when we need to access ops, we can only open-code it.
Huacai
Quoted 41 lines
If adding
has_initis redundant, maybe we can sets->pch_pic_basewithINVALID_GPAinkvm_pch_pic_create()or some other methods. However I think directly accessingkvm_io_device::opsis not a good method, no other architectures do in such way.Regards
Bibo MaoQuoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Huacai Chen
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
In reply to Tao CuiQuoted 52 lines · Tao Cui
Hi, Huacai.
Quoted 39 lines · Huacai Chen
Hi, Tao,
Quoted 36 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ Quoted 9 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 7 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
Quoted 8 lines
Thanks for the suggestion.
I agree that a separate
has_initflag can be unnecessary here. v3 usess->device.opsto track whether the device has already been initialized.The other changes are kept:
pch_pic_baseis updated only after successful bus registration, and the original registration error is returned instead of-EFAULT.
The return value changes can be kept, but why keep the pch_pic_base updating?
Huacai
Quoted 49 lines
Thanks,
TaoQuoted 44 lines
arch/loongarch/kvm/intc/pch_pic.c +3
-0@@ -281,6 +281,9 @@ 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; +if (s->device->ops) +return -EEXIST ; + s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Bibo Mao
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
In reply to Huacai ChenQuoted 70 lines · Huacai Chen
Quoted 62 lines · Bibo Mao
Quoted 58 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -281,6 +281,9 @@ 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;Quoted 5 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0+if (s->device->ops) +return -EEXIST ;This can work, however I think that it is not a good idea to access internal structure field about
kvm_io_device. If so, there is no use about APIkvm_iodevice_init(), justs->device->ops=&kvm_pch_pic_opsis ok.I'm a little not agree. :)
Quoted 7 lines
I think
kvm_iodevice_init()is designed to do more work rather than just set the ops (though it just set the ops now), otherwise its name should bekvm_iodevice_set_ops().In addition, even if
kvm_iodevice_init()is really a setter, there is no getter for the ops, so when we need to access ops, we can only open-code it.
if so, you can try to add kvm_iodevice_get_ops API and check the response of KVM community.
Regards
Bibo Mao
Quoted 45 lines
Huacai
Quoted 41 lines
If adding
has_initis redundant, maybe we can sets->pch_pic_basewithINVALID_GPAinkvm_pch_pic_create()or some other methods. However I think directly accessingkvm_io_device::opsis not a good method, no other architectures do in such way.Regards
Bibo MaoQuoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Huacai Chen
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
Quoted 78 lines · Bibo Mao
Quoted 74 lines · Huacai Chen
Quoted 62 lines · Bibo Mao
Quoted 58 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -281,6 +281,9 @@ 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;Quoted 5 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0+if (s->device->ops) +return -EEXIST ;This can work, however I think that it is not a good idea to access internal structure field about
kvm_io_device. If so, there is no use about APIkvm_iodevice_init(), justs->device->ops=&kvm_pch_pic_opsis ok.I'm a little not agree. :)
I think
kvm_iodevice_init()is designed to do more work rather than just set the ops (though it just set the ops now), otherwise its name should bekvm_iodevice_set_ops().
Quoted 5 lines
Quoted 3 lines
In addition, even if
kvm_iodevice_init()is really a setter, there is no getter for the ops, so when we need to access ops, we can only open-code it.if so, you can try to add
kvm_iodevice_get_opsAPI and check the response of KVM community.
There is not a setter, so I don't think a getter is necessary.
Moreover, you said "no other architectures directly access ops", but in fact, __vgic_doorbell_to_its() from arch/arm64/kvm/vgic/vgic-its.c directly accesses ops.
Huacai
Quoted 50 lines
Regards
Bibo MaoQuoted 45 lines
Huacai
Quoted 41 lines
If adding
has_initis redundant, maybe we can sets->pch_pic_basewithINVALID_GPAinkvm_pch_pic_create()or some other methods. However I think directly accessingkvm_io_device::opsis not a good method, no other architectures do in such way.Regards
Bibo MaoQuoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Bibo Mao
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
Quoted 79 lines · Huacai Chen
Quoted 78 lines · Bibo Mao
Quoted 74 lines · Huacai Chen
Quoted 62 lines · Bibo Mao
Quoted 58 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -281,6 +281,9 @@ 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;Quoted 5 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0+if (s->device->ops) +return -EEXIST ;This can work, however I think that it is not a good idea to access internal structure field about
kvm_io_device. If so, there is no use about APIkvm_iodevice_init(), justs->device->ops=&kvm_pch_pic_opsis ok.I'm a little not agree. :)
I think
kvm_iodevice_init()is designed to do more work rather than just set the ops (though it just set the ops now), otherwise its name should bekvm_iodevice_set_ops().
Quoted 6 lines
Quoted 5 lines
Quoted 3 lines
In addition, even if
kvm_iodevice_init()is really a setter, there is no getter for the ops, so when we need to access ops, we can only open-code it.if so, you can try to add
kvm_iodevice_get_opsAPI and check the response of KVM community.There is not a setter, so I don't think a getter is necessary.
why setter is necessary, getter is not necessary.
Quoted 4 lines
Moreover, you said "no other architectures directly access ops", but in fact,
__vgic_doorbell_to_its()fromarch/arm64/kvm/vgic/vgic-its.cdirectly accesses ops.
If it is used by others, I have no objection any more. But for me I never write such code.
Quoted 54 lines
Huacai
Quoted 50 lines
Regards
Bibo MaoQuoted 45 lines
Huacai
Quoted 41 lines
If adding
has_initis redundant, maybe we can sets->pch_pic_basewithINVALID_GPAinkvm_pch_pic_create()or some other methods. However I think directly accessingkvm_io_device::opsis not a good method, no other architectures do in such way.Regards
Bibo MaoQuoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Huacai Chen
maintainerRe: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
Quoted 88 lines · Bibo Mao
Quoted 84 lines · Huacai Chen
Quoted 83 lines · Bibo Mao
Quoted 70 lines · Huacai Chen
Quoted 62 lines · Bibo Mao
Quoted 58 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -281,6 +281,9 @@ 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;Quoted 5 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0+if (s->device->ops) +return -EEXIST ;This can work, however I think that it is not a good idea to access internal structure field about
kvm_io_device. If so, there is no use about APIkvm_iodevice_init(), justs->device->ops=&kvm_pch_pic_opsis ok.I'm a little not agree. :)
Quoted 7 lines
I think
kvm_iodevice_init()is designed to do more work rather than just set the ops (though it just set the ops now), otherwise its name should bekvm_iodevice_set_ops().In addition, even if
kvm_iodevice_init()is really a setter, there is no getter for the ops, so when we need to access ops, we can only open-code it.if so, you can try to add
kvm_iodevice_get_opsAPI and check the response of KVM community.
Quoted 8 lines
Quoted 1 line
There is not a setter, so I don't think a getter is necessary.
why setter is necessary, getter is not necessary.
Quoted 4 lines
Moreover, you said "no other architectures directly access ops", but in fact,
__vgic_doorbell_to_its()fromarch/arm64/kvm/vgic/vgic-its.cdirectly accesses ops.If it is used by others, I have no objection any more. But for me I never write such code.
OK, Tao, please update your code, drop the 2nd patch and simplify the 4th one. Thanks.
Huacai
Quoted 56 lines
Quoted 55 lines
Huacai
Quoted 50 lines
Regards
Bibo MaoQuoted 45 lines
Huacai
Quoted 41 lines
If adding
has_initis redundant, maybe we can sets->pch_pic_basewithINVALID_GPAinkvm_pch_pic_create()or some other methods. However I think directly accessingkvm_io_device::opsis not a good method, no other architectures do in such way.Regards
Bibo MaoQuoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0
Re: [PATCH v2 4/4] LoongArch: KVM: Reject repeated PCH-PIC CTRL_INIT
Hi, Huacai
Quoted 92 lines · Huacai Chen
Quoted 91 lines · Bibo Mao
Quoted 79 lines · Huacai Chen
Quoted 78 lines · Bibo Mao
Quoted 74 lines · Huacai Chen
Quoted 62 lines · Bibo Mao
Quoted 58 lines · Huacai Chen
Hi, Tao,
Quoted 41 lines · Tao Cui
From: Tao Cui <[email protected]>
KVM_DEV_LOONGARCH_PCH_PIC_CTRL_INIThas no guard against repeated
invocation: every call overwritespch_pic_baseand registers the samekvm_io_deviceon the MMIO bus at the new address, whilekvm_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, tracking the state with ahas_initflag so the check and the MMIO base update are atomic underslots_lock. The base is only committed after a successful bus registration, and the real registration error is propagated instead of being replaced with-EFAULT.Fixes d206d9514873 ("LoongArch: KVM: Add PCHPIC user mode read and write functions")Signed-off-by Tao Cui <[email protected]>--- arch/loongarch/include/asm/kvm_pch_pic.h | 1 + arch/loongarch/kvm/intc/pch_pic.c | 12 ++++++++++-- 2 files changed, 11 insertions(+), 2 deletions(-)arch/loongarch/include/asm/kvm_pch_pic.h +1
-0@@ -53,6 +53,7 @@ struct loongarch_pch_pic {spinlock_t lock;struct kvm *kvm;struct kvm_io_device device; +bool has_init;union pch_pic_id id;uint64_t mask;/* 1:disable irq, 0:enable irq */ uint64_t htmsi_en;/* 1:msi */ arch/loongarch/kvm/intc/pch_pic.c +0
-0arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device *dev, u64 addr) · line 283
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -282,16 +282,24 @@ 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;Why so complicated? The below is enough, no?
arch/loongarch/kvm/intc/pch_pic.c +0
-0@@ -281,6 +281,9 @@ 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;Quoted 5 lines
arch/loongarch/kvm/intc/pch_pic.c static int kvm_pch_pic_init(struct kvm_device · line 286
Quoted 2 lines
arch/loongarch/kvm/intc/pch_pic.c +2
-0+if (s->device->ops) +return -EEXIST ;This can work, however I think that it is not a good idea to access internal structure field about
kvm_io_device. If so, there is no use about APIkvm_iodevice_init(), justs->device->ops=&kvm_pch_pic_opsis ok.I'm a little not agree. :)
I think
kvm_iodevice_init()is designed to do more work rather than just set the ops (though it just set the ops now), otherwise its name should bekvm_iodevice_set_ops().Quoted 6 lines
Quoted 5 lines
Quoted 3 lines
In addition, even if
kvm_iodevice_init()is really a setter, there is no getter for the ops, so when we need to access ops, we can only open-code it.if so, you can try to add
kvm_iodevice_get_opsAPI and check the response of KVM community.There is not a setter, so I don't think a getter is necessary.
why setter is necessary, getter is not necessary.
Quoted 1 line
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Moreover, you said "no other architectures directly access ops", but in fact,
__vgic_doorbell_to_its()fromarch/arm64/kvm/vgic/vgic-its.cdirectly accesses ops.If it is used by others, I have no objection any more. But for me I never write such code.
OK, Tao, please update your code, drop the 2nd patch and simplify the 4th one. Thanks.
I'll send v3 shortly. It drops the dmsintc single-load patch, and the CTRL_INIT patch now checks s->device.ops for repeated initialization instead of the separate has_init flag. pch_pic_base is still updated only after successful bus registration, and the original registration error is returned instead of -EFAULT.
Thanks for the review.
Tao
Quoted 58 lines
Huacai
Quoted 56 lines
Quoted 55 lines
Huacai
Quoted 50 lines
Regards
Bibo MaoQuoted 45 lines
Huacai
Quoted 41 lines
If adding
has_initis redundant, maybe we can sets->pch_pic_basewithINVALID_GPAinkvm_pch_pic_create()or some other methods. However I think directly accessingkvm_io_device::opsis not a good method, no other architectures do in such way.Regards
Bibo MaoQuoted 32 lines
arch/loongarch/kvm/intc/pch_pic.c +1
-0+ s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */Quoted 27 lines
arch/loongarch/kvm/intc/pch_pic.c +10
-3- s->pch_pic_base = addr; device = &s->device;/* init device by pch pic writing and reading ops */ kvm_iodevice_init (device, &kvm_pch_pic_ops);mutex_lock (&kvm->slots_lock); +if (s->has_init) { + ret = -EEXIST ; + goto out; + }/* register pch pic device */ ret =kvm_io_bus_register_dev (kvm,KVM_MMIO_BUS , addr,PCH_PIC_SIZE , device); +if (!ret) { + s->pch_pic_base = addr; + s->has_init = true; + } +out:mutex_unlock (&kvm->slots_lock); -return (ret <0 ) ? -EFAULT :0 ; +return ret; }/* used by user space to get or set pch pic registers */ --2.43.0