Skip to contentSkip to conversation

KVM: arm64: ptdump: Initialize parser_state before pgtable walk

PATCH Applied Zenghui Yu 2 participants 2 messages Mar 28

Subsystems (2)
arm64 KVM arm64
lore
Share & tools
lore

Zenghui Yu

author
#0 lore raw

PATCH 1/1 KVM: arm64: ptdump: Initialize parser_state before pgtable walk

From: "Zenghui Yu (Huawei)" <[email protected]>

If we go through the "need a bigger buffer" path in seq_read_iter(), which is likely to happen as we're dumping page tables, we will pass the populated-by-last-run st::parser_state to kvm_pgtable_walk()/kvm_ptdump_visitor(). As a result, the output of stage2_page_tables on my box looks like

0x0000000240000000-0x0000000000000000 17179869175G 1
0x0000000000000000-0x0000000000200000 2M 2 R px ux AF BLK
0x0000000000200000-0x0000000040000000 1022M 2
0x0000000040000000-0x0000000040200000 2M 2 R W PXNUXN AF BLK
[...]

Fix it by always initializing st::parser_state before starting a new pgtable walk.

Besides, remove st::range as it's not used by note_page(); remove the explicit initialization of parser_state::start_address as it will be initialized in note_page() anyway.

Signed-off-by Zenghui Yu (Huawei) <[email protected]>
---
 arch/arm64/kvm/ptdump.c | 20 +++++++-------------
 1 file changed, 7 insertions(+), 13 deletions(-)
arch/arm64/kvm/ptdump.c +7 -13
index 6a8836207a79..5dfd526c6dc9 100644
@@ -22,7 +22,6 @@ struct kvm_ptdump_guest_state {
 	struct ptdump_pg_state	parser_state;
 	struct addr_marker	ipa_marker[MARKERS_LEN];
 	struct ptdump_pg_level	level[KVM_PGTABLE_MAX_LEVELS];
-	struct ptdump_range	range[MARKERS_LEN];
 };
 
 static const struct ptdump_prot_bits stage2_pte_bits[] = {

@@ -131,16 +130,7 @@ static struct kvm_ptdump_guest_state *kvm_ptdump_parser_create(struct kvm *kvm)
 
 	st->ipa_marker[0].name		= "Guest IPA";
 	st->ipa_marker[1].start_address = BIT(pgtable->ia_bits);
-	st->range[0].end		= BIT(pgtable->ia_bits);
-
 	st->kvm				= kvm;
-	st->parser_state = (struct ptdump_pg_state) {
-		.marker		= &st->ipa_marker[0],
-		.level		= -1,
-		.pg_level	= &st->level[0],
-		.ptdump.range	= &st->range[0],
-		.start_address	= 0,
-	};
 
 	return st;
 }

@@ -151,14 +141,18 @@ static int kvm_ptdump_guest_show(struct seq_file *m, void *unused)
 	struct kvm_ptdump_guest_state *st = m->private;
 	struct kvm *kvm = st->kvm;
 	struct kvm_s2_mmu *mmu = &kvm->arch.mmu;
-	struct ptdump_pg_state *parser_state = &st->parser_state;
 	struct kvm_pgtable_walker walker = (struct kvm_pgtable_walker) {
 		.cb	= kvm_ptdump_visitor,
-		.arg	= parser_state,
+		.arg	= &st->parser_state,
 		.flags	= KVM_PGTABLE_WALK_LEAF,
 	};
 
-	parser_state->seq = m;
+	st->parser_state = (struct ptdump_pg_state) {
+		.marker		= &st->ipa_marker[0],
+		.level		= -1,
+		.pg_level	= &st->level[0],
+		.seq		= m,
+	};
 
 	write_lock(&kvm->mmu_lock);
 	ret = kvm_pgtable_walk(mmu->pgt, 0, BIT(mmu->pgt->ia_bits), &walker);
Signature · 1 line
2.53.0
maintainer Applied
#1 lore raw
Quoted 6 lines · Zenghui Yu

If we go through the "need a bigger buffer" path in seq_read_iter(), which is likely to happen as we're dumping page tables, we will pass the populated-by-last-run st::parser_state to kvm_pgtable_walk()/kvm_ptdump_visitor(). As a result, the output of stage2_page_tables on my box looks like

Quoted 7 lines

0x0000000240000000-0x0000000000000000 17179869175G 1
0x0000000000000000-0x0000000000200000 2M 2 R px ux AF BLK
0x0000000000200000-0x0000000040000000 1022M 2
0x0000000040000000-0x0000000040200000 2M 2 R W PXNUXN AF BLK
[...]

[...]

Applied to next, thanks!

[1/1] KVM: arm64: ptdump: Initialize parser_state before pgtable walk

      commit: 570428601ba506e76c265a65626524ef3c5cbc04

Cheers,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.