Skip to contentSkip to conversation

migration: Fix migration crash when target psize larger than host

PATCH In review Peter Xu 3 participants 3 messages Jan 20, 2023 – Jan 30, 2023

lore
Share & tools
lore

Peter Xu

author
#0 lore raw

PATCH 1/1 migration: Fix migration crash when target psize larger than host

Commit d9e474ea56 overlooked the case where the target psize is even larger than the host psize. One example is Alpha has 8K page size and migration will start to crash the source QEMU when running Alpha migration on x86.

Fix it by detecting that case and set host start/end just to cover the single page to be migrated.

This will slightly optimize the common case where host psize equals to guest psize so we don't even need to do the roundups, but that's trivial.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1456

Fixes d9e474ea56 ("migration: Teach PSS about host page")
---
 migration/ram.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
migration/ram.c +19 -2
index 334309f1c6..68a45338e3 100644
@@ -2319,8 +2319,25 @@ static void pss_host_page_prepare(PageSearchStatus *pss)
     size_t guest_pfns = qemu_ram_pagesize(pss->block) >> TARGET_PAGE_BITS;
 
     pss->host_page_sending = true;
-    pss->host_page_start = ROUND_DOWN(pss->page, guest_pfns);
-    pss->host_page_end = ROUND_UP(pss->page + 1, guest_pfns);
+    if (guest_pfns <= 1) {
+        /*
+         * This covers both when guest psize == host psize, or when guest
+         * has larger psize than the host (guest_pfns==0).
+         *
+         * For the latter, we always send one whole guest page per
+         * iteration of the host page (example: an Alpha VM on x86 host
+         * will have guest psize 8K while host psize 4K).
+         */
+        pss->host_page_start = pss->page;
+        pss->host_page_end = pss->page + 1;
+    } else {
+        /*
+         * The host page spans over multiple guest pages, we send them
+         * within the same host page iteration.
+         */
+        pss->host_page_start = ROUND_DOWN(pss->page, guest_pfns);
+        pss->host_page_end = ROUND_UP(pss->page + 1, guest_pfns);
+    }
 }
 
 /*
Signature · 1 line
2.37.3
Tested-by on v1 1/1
#1 lore raw
Quoted 10 lines · Peter Xu

Commit d9e474ea56 overlooked the case where the target psize is even larger than the host psize. One example is Alpha has 8K page size and migration will start to crash the source QEMU when running Alpha migration on x86.

Fix it by detecting that case and set host start/end just to cover the single page to be migrated.

This will slightly optimize the common case where host psize equals to guest psize so we don't even need to do the roundups, but that's trivial.

Quoted 8 lines

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1456

Fixes d9e474ea56 ("migration: Teach PSS about host page")
---
  migration/ram.c | 21 +++++++++++++++++++--
  1 file changed, 19 insertions(+), 2 deletions(-)

Thanks, I've checked that this fixes the issue for me, indeed!

Reviewed-by on v1 1/1
#2 lore raw
In reply to Peter Xu
Quoted 7 lines · Peter Xu

Commit d9e474ea56 overlooked the case where the target psize is even larger than the host psize. One example is Alpha has 8K page size and migration will start to crash the source QEMU when running Alpha migration on x86.

Fix it by detecting that case and set host start/end just to cover the single page to be migrated.

Quoted 8 lines

This will slightly optimize the common case where host psize equals to guest psize so we don't even need to do the roundups, but that's trivial.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1456

Fixes d9e474ea56 ("migration: Teach PSS about host page")

queued