Skip to contentSkip to conversation

Fix a wrong value passed to __find_vmap_area()

PATCH In review Uladzislau Rezki (Sony) 4 participants 5 messages Jan 11, 2024 – Jan 17, 2024

Subsystems (2)
lore
Share & tools
lore

Uladzislau Rezki (Sony)

author maintainer
#0 lore raw

PATCH 1/1 Fix a wrong value passed to __find_vmap_area()

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Fixes e88b85b81d5 ("mm: vmalloc: offload free_vmap_area_lock lock")
---
 mm/vmalloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
mm/vmalloc.c +1 -1
index e30dabf68263..43a6608e1397 100644
@@ -4688,7 +4688,7 @@ bool vmalloc_dump_obj(void *object)
 	vn = addr_to_node((unsigned long)objp);
 
 	if (spin_trylock(&vn->busy.lock)) {
-		va = __find_vmap_area(addr, &vn->busy.root);
+		va = __find_vmap_area((unsigned long)objp, &vn->busy.root);
 
 		if (va && va->vm) {
 			addr = (unsigned long)va->vm->addr;
Signature · 1 line
2.39.2
#1 lore raw
Quoted 6 lines · Uladzislau Rezki (Sony)

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Fixes e88b85b81d5 ("mm: vmalloc: offload free_vmap_area_lock lock")

I am not sure that e88b85b81d5 introduced this. Based on the diff, I think it was actually commit 86817057732a ("mm: vmalloc: remove global vmap_area_root rb-tree"). Does not really matter for the patch text but I think it does matter for what change this gets squashed into.

Quoted 22 lines
---
 mm/vmalloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
mm/vmalloc.c +1 -1
index e30dabf68263..43a6608e1397 100644
@@ -4688,7 +4688,7 @@ bool vmalloc_dump_obj(void *object)
 	vn = addr_to_node((unsigned long)objp);
 
 	if (spin_trylock(&vn->busy.lock)) {
-		va = __find_vmap_area(addr, &vn->busy.root);
+		va = __find_vmap_area((unsigned long)objp, &vn->busy.root);
 
 		if (va && va->vm) {
 			addr = (unsigned long)va->vm->addr;
Signature · 1 line
2.39.2
author maintainer
#2 lore raw
Quoted 5 lines · Nathan Chancellor
Quoted 4 lines · Uladzislau Rezki (Sony)

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Quoted 8 lines
Quoted 2 lines
Fixes e88b85b81d5 ("mm: vmalloc: offload free_vmap_area_lock lock")

I am not sure that e88b85b81d5 introduced this. Based on the diff, I think it was actually commit 86817057732a ("mm: vmalloc: remove global vmap_area_root rb-tree"). Does not really matter for the patch text but I think it does matter for what change this gets squashed into.

Right you are! Below is updated version:

<snip>
From 98f1fd2d3913f3b7bcbe49785a78f67999151f1c Mon Sep 17 00:00:00 2001
From: "Uladzislau Rezki (Sony)" <[email protected]>
Date: Thu, 11 Jan 2024 13:04:07 +0100
Subject: [PATCH 1/2] mm: vmalloc: Fix a wrong value passed to
__find_vmap_area()

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Fixes 86817057732a ("mm: vmalloc: remove global vmap_area_root rb-tree")
---
 mm/vmalloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
mm/vmalloc.c +1 -1
index e30dabf68263..43a6608e1397 100644
@@ -4688,7 +4688,7 @@ bool vmalloc_dump_obj(void *object)
 	vn = addr_to_node((unsigned long)objp);
 
 	if (spin_trylock(&vn->busy.lock)) {
-		va = __find_vmap_area(addr, &vn->busy.root);
+		va = __find_vmap_area((unsigned long)objp, &vn->busy.root);
 
 		if (va && va->vm) {
 			addr = (unsigned long)va->vm->addr;
Signature · 5 lines
2.39.2
<snip>

--
Uladzislau Rezki
Reviewed-by
#3 lore raw
Quoted 23 lines · Uladzislau Rezki
Quoted 5 lines · Nathan Chancellor
Quoted 4 lines · Uladzislau Rezki (Sony)

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Quoted 8 lines
Quoted 2 lines
Fixes e88b85b81d5 ("mm: vmalloc: offload free_vmap_area_lock lock")

I am not sure that e88b85b81d5 introduced this. Based on the diff, I think it was actually commit 86817057732a ("mm: vmalloc: remove global vmap_area_root rb-tree"). Does not really matter for the patch text but I think it does matter for what change this gets squashed into.

Right you are! Below is updated version:

<snip>
From 98f1fd2d3913f3b7bcbe49785a78f67999151f1c Mon Sep 17 00:00:00 2001
From: "Uladzislau Rezki (Sony)" <[email protected]>
Date: Thu, 11 Jan 2024 13:04:07 +0100
Subject: [PATCH 1/2] mm: vmalloc: Fix a wrong value passed to
__find_vmap_area()

Quoted 6 lines

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Fixes 86817057732a ("mm: vmalloc: remove global vmap_area_root rb-tree")

I know the commits are likely to get squashed/messed with (this is now d1d9bdd672c4 in my mm-unstable tree), will this get corrected in the commit message also? Slightly tricky one.

Perhaps a note for Andrew unless his scripts do this already - please update this to wherever "mm: vmalloc: remove global vmap_area_root rb-tree" lands?

Quoted 23 lines
---
 mm/vmalloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
mm/vmalloc.c +1 -2
index e30dabf68263..43a6608e1397 100644
@@ -4688,7 +4688,7 @@ bool vmalloc_dump_obj(void *object)
 	vn = addr_to_node((unsigned long)objp);
 
 	if (spin_trylock(&vn->busy.lock)) {
-		va = __find_vmap_area(addr, &vn->busy.root);
+		va = __find_vmap_area((unsigned long)objp, &vn->busy.root);
 
 		if (va && va->vm) {
 			addr = (unsigned long)va->vm->addr;
--

2.39.2
<snip>

Quoted 2 lines

--
Uladzislau Rezki

Otherwise looks good to me. I have some style comments but perhaps better for me to address on the actual patch series! Feel free to add:

maintainer
#4 lore raw
Quoted 7 lines · On Tue, 16 Jan 2024 22:13:17 +0000 Lorenzo Stoakes <[email protected]> wrote:
Quoted 6 lines

There was a type in the vmalloc_dump_obj() function. Instead of passing a real address which is "objp" an "addr" was used what is wrong and not initialized.

Fixes 86817057732a ("mm: vmalloc: remove global vmap_area_root rb-tree")
Quoted 7 lines

I know the commits are likely to get squashed/messed with (this is now d1d9bdd672c4 in my mm-unstable tree), will this get corrected in the commit message also? Slightly tricky one.

Perhaps a note for Andrew unless his scripts do this already - please update this to wherever "mm: vmalloc: remove global vmap_area_root rb-tree" lands?

Yep, I'll scrunch together

mm-vmalloc-remove-global-vmap_area_root-rb-tree.patch mm-vmalloc-remove-global-vmap_area_root-rb-tree-fix.patch mm-vmalloc-remove-global-vmap_area_root-rb-tree-fix-2.patch

before merging it all into mm-stable and shall tidy up the changelog trail.