On Sat, Feb 01, 2025 at 09:57:23AM +0000, “William Roche wrote:
> From: William Roche <william.roche@xxxxxxxxxx>
>
> In case of a large page impacted by a memory error, provide an
> information about the impacted large page before the memory
> error injection message.
>
> This message would also appear on ras enabled ARM platforms, with
> the introduction of an x86 similar error injection message.
>
> In the case of a large page impacted, we now report:
> Memory Error on large page from <backend>:<address>+<fd_offset> +<page_size>
>
> The +<fd_offset> information is only provided with a file backend.
>
> Signed-off-by: William Roche <william.roche@xxxxxxxxxx>
This is still pretty kvm / arch relevant patch that needs some reviews.
I wonder do we really need this - we could fetch ramblock mapping
(e.g. hwaddr -> HVA) via HMP "info ramblock", and also dmesg shows process
ID + VA. IIUC we have all below info already as long as we do some math
based on above. Would that work too?
> ---
> accel/kvm/kvm-all.c | 18 ++++++++++++++++++
> include/exec/cpu-common.h | 10 ++++++++++
> system/physmem.c | 22 ++++++++++++++++++++++
> target/arm/kvm.c | 3 +++
> 4 files changed, 53 insertions(+)
>
> diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
> index f89568bfa3..9a0d970ce1 100644
> --- a/accel/kvm/kvm-all.c
> +++ b/accel/kvm/kvm-all.c
> @@ -1296,6 +1296,24 @@ static void kvm_unpoison_all(void *param)
> void kvm_hwpoison_page_add(ram_addr_t ram_addr)
> {
> HWPoisonPage *page;
> + struct RAMBlockInfo rb_info;
> +
> + if (qemu_ram_block_info_from_addr(ram_addr, &rb_info)) {
> + size_t ps = rb_info.page_size;
> +
> + if (ps > TARGET_PAGE_SIZE) {
> + uint64_t offset = QEMU_ALIGN_DOWN(ram_addr - rb_info.offset, ps);
> +
> + if (rb_info.fd >= 0) {
> + error_report("Memory Error on large page from %s:%" PRIx64
> + "+%" PRIx64 " +%zx", rb_info.idstr, offset,
> + rb_info.fd_offset, ps);
> + } else {
> + error_report("Memory Error on large page from %s:%" PRIx64
> + " +%zx", rb_info.idstr, offset, ps);
> + }
> + }
> + }
>
> QLIST_FOREACH(page, &hwpoison_page_list, list) {
> if (page->ram_addr == ram_addr) {
> diff --git a/include/exec/cpu-common.h b/include/exec/cpu-common.h
> index 3771b2130c..190bd4f34a 100644
> --- a/include/exec/cpu-common.h
> +++ b/include/exec/cpu-common.h
> @@ -110,6 +110,16 @@ int qemu_ram_get_fd(RAMBlock *rb);
> size_t qemu_ram_pagesize(RAMBlock *block);
> size_t qemu_ram_pagesize_largest(void);
>
> +struct RAMBlockInfo {
> + char idstr[256];
> + ram_addr_t offset;
> + int fd;
> + uint64_t fd_offset;
> + size_t page_size;
> +};
> +bool qemu_ram_block_info_from_addr(ram_addr_t ram_addr,
> + struct RAMBlockInfo *block);
> +
> /**
> * cpu_address_space_init:
> * @cpu: CPU to add this address space to
> diff --git a/system/physmem.c b/system/physmem.c
> index e8ff930bc9..686f569270 100644
> --- a/system/physmem.c
> +++ b/system/physmem.c
> @@ -1678,6 +1678,28 @@ size_t qemu_ram_pagesize_largest(void)
> return largest;
> }
>
> +/* Copy RAMBlock information associated to the given ram_addr location */
> +bool qemu_ram_block_info_from_addr(ram_addr_t ram_addr,
> + struct RAMBlockInfo *b_info)
> +{
> + RAMBlock *rb;
> +
> + assert(b_info);
> +
> + RCU_READ_LOCK_GUARD();
> + rb = qemu_get_ram_block(ram_addr);
> + if (!rb) {
> + return false;
> + }
> +
> + pstrcat(b_info->idstr, sizeof(b_info->idstr), rb->idstr);
> + b_info->offset = rb->offset;
> + b_info->fd = rb->fd;
> + b_info->fd_offset = rb->fd_offset;
> + b_info->page_size = rb->page_size;
> + return true;
> +}
> +
> static int memory_try_enable_merging(void *addr, size_t len)
> {
> if (!machine_mem_merge(current_machine)) {
> diff --git a/target/arm/kvm.c b/target/arm/kvm.c
> index da30bdbb23..d9dedc6d74 100644
> --- a/target/arm/kvm.c
> +++ b/target/arm/kvm.c
> @@ -2389,6 +2389,9 @@ void kvm_arch_on_sigbus_vcpu(CPUState *c, int code, void *addr)
> kvm_cpu_synchronize_state(c);
> if (!acpi_ghes_memory_errors(ACPI_HEST_SRC_ID_SEA, paddr)) {
> kvm_inject_arm_sea(c);
> + error_report("Guest Memory Error at QEMU addr %p and "
> + "GUEST addr 0x%" HWADDR_PRIx " of type %s injected",
> + addr, paddr, "BUS_MCEERR_AR");
> } else {
> error_report("failed to record the error");
> abort();
> --
> 2.43.5
>
--
Peter Xu
