qemu exits during reset with log:
qemu-system-x86_64: Could not remap addr: 1000@22001000
Currently, after MCE on RAM of a guest, qemu records a ram_addr only,
remaps this address with a fixed size(TARGET_PAGE_SIZE) during reset.
In the hugetlbfs scenario, mmap(addr...) needs page_size aligned
address and correct size. Unaligned address leads mmap to fail.
What's more, hitting MCE on RAM of a guest, qemu records this address
and try to fix it during reset, this should be a common logic. So
remove kvm_hwpoison_page_add from architecture dependent code, record
this in SIGBUS handler instead. Finally poisoning/unpoisoning a page
gets static in kvm-all.c,
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
---
accel/kvm/kvm-all.c | 47 ++++++++++++++++++++++++++++++----------
include/sysemu/kvm_int.h | 12 ----------
target/arm/kvm64.c | 1 -
target/i386/kvm/kvm.c | 1 -
4 files changed, 36 insertions(+), 25 deletions(-)
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index 5f1377ca04..2a91c5a461 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -1167,11 +1167,14 @@ int kvm_vm_check_extension(KVMState *s, unsigned int
extension)
return ret;
}
+#ifdef KVM_HAVE_MCE_INJECTION
typedef struct HWPoisonPage {
ram_addr_t ram_addr;
+ size_t page_size; /* normal page or hugeTLB page? */
QLIST_ENTRY(HWPoisonPage) list;
} HWPoisonPage;
+/* hwpoison_page_list stores the poisoned pages, unpoison them during reset */
static QLIST_HEAD(, HWPoisonPage) hwpoison_page_list =
QLIST_HEAD_INITIALIZER(hwpoison_page_list);
@@ -1181,25 +1184,48 @@ static void kvm_unpoison_all(void *param)
QLIST_FOREACH_SAFE(page, &hwpoison_page_list, list, next_page) {
QLIST_REMOVE(page, list);
- qemu_ram_remap(page->ram_addr, TARGET_PAGE_SIZE);
+ qemu_ram_remap(page->ram_addr, page->page_size);
g_free(page);
}
}
-void kvm_hwpoison_page_add(ram_addr_t ram_addr)
+static void kvm_hwpoison_page_add(CPUState *cpu, int sigbus_code, void *addr)
{
HWPoisonPage *page;
+ ram_addr_t ram_addr, align_ram_addr;
+ ram_addr_t offset;
+ hwaddr paddr;
+ size_t page_size;
+
+ assert(sigbus_code == BUS_MCEERR_AR || sigbus_code == BUS_MCEERR_AO);
+ ram_addr = qemu_ram_addr_from_host(addr);
+ if (ram_addr == RAM_ADDR_INVALID ||
+ !kvm_physical_memory_addr_from_host(cpu->kvm_state, addr, &paddr)) {
+ /* only deal with valid guest RAM here */
+ return;
+ }
+ /* get page size of RAM block, test it's a normal page or huge page */
+ page_size = qemu_ram_block_from_host(addr, false, &offset)->page_size;
+ align_ram_addr = QEMU_ALIGN_DOWN(ram_addr, page_size);
QLIST_FOREACH(page, &hwpoison_page_list, list) {
- if (page->ram_addr == ram_addr) {
+ if (page->ram_addr == align_ram_addr) {
+ assert(page->page_size == page_size);
return;
}
}
- page = g_new(HWPoisonPage, 1);
- page->ram_addr = ram_addr;
+
+ page = g_new0(HWPoisonPage, 1);
+ page->ram_addr = align_ram_addr;
+ page->page_size = page_size;
QLIST_INSERT_HEAD(&hwpoison_page_list, page, list);
}
+static __thread void *pending_sigbus_addr;
+static __thread int pending_sigbus_code;
+static __thread bool have_sigbus_pending;
+#endif
+
static uint32_t adjust_ioeventfd_endianness(uint32_t val, uint32_t size)
{
#if defined(HOST_WORDS_BIGENDIAN) != defined(TARGET_WORDS_BIGENDIAN)
@@ -2601,7 +2627,9 @@ static int kvm_init(MachineState *ms)
s->kernel_irqchip_split = mc->default_kernel_irqchip_split ?
ON_OFF_AUTO_ON : ON_OFF_AUTO_OFF;
}
+#if defined KVM_HAVE_MCE_INJECTION
qemu_register_reset(kvm_unpoison_all, NULL);
+#endif
if (s->kernel_irqchip_allowed) {
kvm_irqchip_create(s);
@@ -2782,12 +2810,6 @@ void kvm_cpu_synchronize_pre_loadvm(CPUState *cpu)
run_on_cpu(cpu, do_kvm_cpu_synchronize_pre_loadvm, RUN_ON_CPU_NULL);
}
-#ifdef KVM_HAVE_MCE_INJECTION
-static __thread void *pending_sigbus_addr;
-static __thread int pending_sigbus_code;
-static __thread bool have_sigbus_pending;
-#endif
-
static void kvm_cpu_kick(CPUState *cpu)
{
qatomic_set(&cpu->kvm_run->immediate_exit, 1);
@@ -2883,6 +2905,8 @@ int kvm_cpu_exec(CPUState *cpu)
#ifdef KVM_HAVE_MCE_INJECTION
if (unlikely(have_sigbus_pending)) {
qemu_mutex_lock_iothread();
+ kvm_hwpoison_page_add(cpu, pending_sigbus_code,
+ pending_sigbus_addr);
kvm_arch_on_sigbus_vcpu(cpu, pending_sigbus_code,
pending_sigbus_addr);
have_sigbus_pending = false;
@@ -3436,6 +3460,7 @@ int kvm_on_sigbus(int code, void *addr)
* we can only get action optional here.
*/
assert(code != BUS_MCEERR_AR);
+ kvm_hwpoison_page_add(first_cpu, code, addr);
kvm_arch_on_sigbus_vcpu(first_cpu, code, addr);
return 0;
#else
diff --git a/include/sysemu/kvm_int.h b/include/sysemu/kvm_int.h
index 1f5487d9b7..52ec8ef99c 100644
--- a/include/sysemu/kvm_int.h
+++ b/include/sysemu/kvm_int.h
@@ -40,16 +40,4 @@ void kvm_memory_listener_register(KVMState *s,
KVMMemoryListener *kml,
AddressSpace *as, int as_id, const char
*name);
void kvm_set_max_memslot_size(hwaddr max_slot_size);
-
-/**
- * kvm_hwpoison_page_add:
- *
- * Parameters:
- * @ram_addr: the address in the RAM for the poisoned page
- *
- * Add a poisoned page to the list
- *
- * Return: None.
- */
-void kvm_hwpoison_page_add(ram_addr_t ram_addr);
#endif
diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c
index ccadfbbe72..a3184eb3d2 100644
--- a/target/arm/kvm64.c
+++ b/target/arm/kvm64.c
@@ -1450,7 +1450,6 @@ void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void
*addr)
ram_addr = qemu_ram_addr_from_host(addr);
if (ram_addr != RAM_ADDR_INVALID &&
kvm_physical_memory_addr_from_host(c->kvm_state, addr, &paddr)) {
- kvm_hwpoison_page_add(ram_addr);
/*
* If this is a BUS_MCEERR_AR, we know we have been called
* synchronously from the vCPU thread, so we can easily
diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
index 9cf8e03669..fb72b349ed 100644
--- a/target/i386/kvm/kvm.c
+++ b/target/i386/kvm/kvm.c
@@ -622,7 +622,6 @@ void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void
*addr)
ram_addr = qemu_ram_addr_from_host(addr);
if (ram_addr != RAM_ADDR_INVALID &&
kvm_physical_memory_addr_from_host(c->kvm_state, addr, &paddr)) {
- kvm_hwpoison_page_add(ram_addr);
kvm_mce_inject(cpu, paddr, code);
/*