Skip to contentSkip to conversation

btrfs: Consider system chunk array size for new SYSTEM chunks

PATCH In review Qu Wenruo 4 participants 10 messages Aug 28, 2019 – Oct 24, 2019 Btrfs file system

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

PATCH 1/2 btrfs: Consider system chunk array size for new SYSTEM chunks

For SYSTEM chunks, despite the regular chunk item size limit, there is another limit due to system chunk array size.

The extra limit is removed in a refactor, so just add it back.

Fixes e3ecdb3fdecf ("btrfs: factor out devs_max setting in __btrfs_alloc_chunk")
---
 fs/btrfs/volumes.c | 1 +
 1 file changed, 1 insertion(+)
fs/btrfs/volumes.c +1 -0
index a447d3ec48d5..8b72d03738d9 100644
@@ -4966,6 +4966,7 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 	} else if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
 		max_stripe_size = SZ_32M;
 		max_chunk_size = 2 * max_stripe_size;
+		devs_max = min_t(int, devs_max, BTRFS_MAX_DEVS_SYS_CHUNK);
 	} else {
 		btrfs_err(info, "invalid chunk type 0x%llx requested",
 		       type);
Signature · 1 line
2.23.0
author
#1 lore raw

PATCH v3 2/2 btrfs: tree-checker: Fix wrong check on max devid

[BUG]
The following script will cause false alert on devid check.
#!/bin/bash

dev1=/dev/test/test
dev2=/dev/test/scratch1
mnt=/mnt/btrfs

umount $dev1 &> /dev/null
umount $dev2 &> /dev/null
umount $mnt &> /dev/null

mkfs.btrfs -f $dev1

mount $dev1 $mnt

  _fail()
  {
          echo "!!! FAILED !!!"
          exit 1
  }

  for ((i = 0; i < 4096; i++)); do
          btrfs dev add -f $dev2 $mnt || _fail
          btrfs dev del $dev1 $mnt || _fail
          dev_tmp=$dev1
          dev1=$dev2
          dev2=$dev_tmp
  done

[CAUSE]
Tree-checker uses BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK() as upper limit for devid.
But we can have devid holes just like above script.

So the check for devid is incorrect and could cause false alert.

[FIX]
Just remove the whole devid check.
We don't have any hard requirement for devid assignment.

Furthermore, even devid get corrupted by bitflip, we still have dev extents verification at mount time, so corrupted data won't sneak into kernel.

Fixes ab4ba2e13346 ("btrfs: tree-checker: Verify dev item")

---
Changelog:
v2:
- Remove devid check completely
As we already have verify_one_dev_extent().
v3:
- Unexport BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK macros
- Update commit message to include the reproducer.

---
 fs/btrfs/tree-checker.c | 8 --------
 fs/btrfs/volumes.c      | 9 +++++++++
 fs/btrfs/volumes.h      | 9 ---------
 3 files changed, 9 insertions(+), 17 deletions(-)
fs/btrfs/tree-checker.c +0 -8
index ccd5706199d7..15d1aa7cef1f 100644
@@ -686,9 +686,7 @@ static void dev_item_err(const struct extent_buffer *eb, int slot,
 static int check_dev_item(struct extent_buffer *leaf,
 			  struct btrfs_key *key, int slot)
 {
-	struct btrfs_fs_info *fs_info = leaf->fs_info;
 	struct btrfs_dev_item *ditem;
-	u64 max_devid = max(BTRFS_MAX_DEVS(fs_info), BTRFS_MAX_DEVS_SYS_CHUNK);
 
 	if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
 		dev_item_err(leaf, slot,
@@ -696,12 +694,6 @@ static int check_dev_item(struct extent_buffer *leaf,
 			     key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
 		return -EUCLEAN;
 	}
-	if (key->offset > max_devid) {
-		dev_item_err(leaf, slot,
-			     "invalid devid: has=%llu expect=[0, %llu]",
-			     key->offset, max_devid);
-		return -EUCLEAN;
-	}
 	ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
 	if (btrfs_device_id(leaf, ditem) != key->offset) {
 		dev_item_err(leaf, slot,
fs/btrfs/volumes.c +9 -0
index 8b72d03738d9..56f751192a6c 100644
@@ -4901,6 +4901,15 @@ static void check_raid56_incompat_flag(struct btrfs_fs_info *info, u64 type)
 	btrfs_set_fs_incompat(info, RAID56);
 }
 
+#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
+			- sizeof(struct btrfs_chunk))		\
+			/ sizeof(struct btrfs_stripe) + 1)
+
+#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
+				- 2 * sizeof(struct btrfs_disk_key)	\
+				- 2 * sizeof(struct btrfs_chunk))	\
+				/ sizeof(struct btrfs_stripe) + 1)
+
 static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 			       u64 start, u64 type)
 {
fs/btrfs/volumes.h +0 -9
index 7f6aa1816409..789f983a98c5 100644
@@ -273,15 +273,6 @@ struct btrfs_fs_devices {
 
 #define BTRFS_BIO_INLINE_CSUM_SIZE	64
 
-#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
-			- sizeof(struct btrfs_chunk))		\
-			/ sizeof(struct btrfs_stripe) + 1)
-
-#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
-				- 2 * sizeof(struct btrfs_disk_key)	\
-				- 2 * sizeof(struct btrfs_chunk))	\
-				/ sizeof(struct btrfs_stripe) + 1)
-
 /*
  * we need the mirror number and stripe index to be passed around
  * the call chain while we are processing end_io (especially errors).
Signature · 1 line
2.23.0
#2 lore raw

Re: [PATCH v3 2/2] btrfs: tree-checker: Fix wrong check on max devid

Quoted 16 lines · Qu Wenruo

[BUG]
The following script will cause false alert on devid check.
#!/bin/bash

dev1=/dev/test/test
dev2=/dev/test/scratch1
mnt=/mnt/btrfs

umount $dev1 &> /dev/null
umount $dev2 &> /dev/null
umount $mnt &> /dev/null

mkfs.btrfs -f $dev1

mount $dev1 $mnt

Quoted 13 lines
  _fail()
  {
          echo "!!! FAILED !!!"
          exit 1
  }

  for ((i = 0; i < 4096; i++)); do
          btrfs dev add -f $dev2 $mnt || _fail
          btrfs dev del $dev1 $mnt || _fail
          dev_tmp=$dev1
          dev1=$dev2
          dev2=$dev_tmp
  done

Instead of putting the script here, can't it be turned into a fstest to ensure we don't regress in the future?

Quoted 101 lines

[CAUSE]
Tree-checker uses BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK() as upper limit for devid.
But we can have devid holes just like above script.

So the check for devid is incorrect and could cause false alert.

[FIX]
Just remove the whole devid check.
We don't have any hard requirement for devid assignment.

Furthermore, even devid get corrupted by bitflip, we still have dev extents verification at mount time, so corrupted data won't sneak into kernel.

Fixes ab4ba2e13346 ("btrfs: tree-checker: Verify dev item")

---
Changelog:
v2:
- Remove devid check completely
As we already have verify_one_dev_extent().
v3:
- Unexport BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK macros
- Update commit message to include the reproducer.

---
 fs/btrfs/tree-checker.c | 8 --------
 fs/btrfs/volumes.c      | 9 +++++++++
 fs/btrfs/volumes.h      | 9 ---------
 3 files changed, 9 insertions(+), 17 deletions(-)
fs/btrfs/tree-checker.c +0 -8
index ccd5706199d7..15d1aa7cef1f 100644
@@ -686,9 +686,7 @@ static void dev_item_err(const struct extent_buffer *eb, int slot,
 static int check_dev_item(struct extent_buffer *leaf,
 			  struct btrfs_key *key, int slot)
 {
-	struct btrfs_fs_info *fs_info = leaf->fs_info;
 	struct btrfs_dev_item *ditem;
-	u64 max_devid = max(BTRFS_MAX_DEVS(fs_info), BTRFS_MAX_DEVS_SYS_CHUNK);
 
 	if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
 		dev_item_err(leaf, slot,
@@ -696,12 +694,6 @@ static int check_dev_item(struct extent_buffer *leaf,
 			     key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
 		return -EUCLEAN;
 	}
-	if (key->offset > max_devid) {
-		dev_item_err(leaf, slot,
-			     "invalid devid: has=%llu expect=[0, %llu]",
-			     key->offset, max_devid);
-		return -EUCLEAN;
-	}
 	ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
 	if (btrfs_device_id(leaf, ditem) != key->offset) {
 		dev_item_err(leaf, slot,
fs/btrfs/volumes.c +9 -0
index 8b72d03738d9..56f751192a6c 100644
@@ -4901,6 +4901,15 @@ static void check_raid56_incompat_flag(struct btrfs_fs_info *info, u64 type)
 	btrfs_set_fs_incompat(info, RAID56);
 }
 
+#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
+			- sizeof(struct btrfs_chunk))		\
+			/ sizeof(struct btrfs_stripe) + 1)
+
+#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
+				- 2 * sizeof(struct btrfs_disk_key)	\
+				- 2 * sizeof(struct btrfs_chunk))	\
+				/ sizeof(struct btrfs_stripe) + 1)
+
 static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 			       u64 start, u64 type)
 {
fs/btrfs/volumes.h +0 -9
index 7f6aa1816409..789f983a98c5 100644
@@ -273,15 +273,6 @@ struct btrfs_fs_devices {
 
 #define BTRFS_BIO_INLINE_CSUM_SIZE	64
 
-#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
-			- sizeof(struct btrfs_chunk))		\
-			/ sizeof(struct btrfs_stripe) + 1)
-
-#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
-				- 2 * sizeof(struct btrfs_disk_key)	\
-				- 2 * sizeof(struct btrfs_chunk))	\
-				/ sizeof(struct btrfs_stripe) + 1)
-
 /*
  * we need the mirror number and stripe index to be passed around
  * the call chain while we are processing end_io (especially errors).
author
#3 lore raw

Re: [PATCH v3 2/2] btrfs: tree-checker: Fix wrong check on max devid

Quoted 33 lines · Nikolay Borisov
Quoted 16 lines · Qu Wenruo

[BUG]
The following script will cause false alert on devid check.
#!/bin/bash

dev1=/dev/test/test
dev2=/dev/test/scratch1
mnt=/mnt/btrfs

umount $dev1 &> /dev/null
umount $dev2 &> /dev/null
umount $mnt &> /dev/null

mkfs.btrfs -f $dev1

mount $dev1 $mnt

Quoted 13 lines
  _fail()
  {
          echo "!!! FAILED !!!"
          exit 1
  }

  for ((i = 0; i < 4096; i++)); do
          btrfs dev add -f $dev2 $mnt || _fail
          btrfs dev del $dev1 $mnt || _fail
          dev_tmp=$dev1
          dev1=$dev2
          dev2=$dev_tmp
  done
Quoted 2 lines

Instead of putting the script here, can't it be turned into a fstest to ensure we don't regress in the future?

Sure, already crafting.

Thanks,
Qu

Quoted 102 lines
Quoted 102 lines

[CAUSE]
Tree-checker uses BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK() as upper limit for devid.
But we can have devid holes just like above script.

So the check for devid is incorrect and could cause false alert.

[FIX]
Just remove the whole devid check.
We don't have any hard requirement for devid assignment.

Furthermore, even devid get corrupted by bitflip, we still have dev extents verification at mount time, so corrupted data won't sneak into kernel.

Fixes ab4ba2e13346 ("btrfs: tree-checker: Verify dev item")

---
Changelog:
v2:
- Remove devid check completely
As we already have verify_one_dev_extent().
v3:
- Unexport BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK macros
- Update commit message to include the reproducer.

---
 fs/btrfs/tree-checker.c | 8 --------
 fs/btrfs/volumes.c      | 9 +++++++++
 fs/btrfs/volumes.h      | 9 ---------
 3 files changed, 9 insertions(+), 17 deletions(-)
fs/btrfs/tree-checker.c +0 -8
index ccd5706199d7..15d1aa7cef1f 100644
@@ -686,9 +686,7 @@ static void dev_item_err(const struct extent_buffer *eb, int slot,
 static int check_dev_item(struct extent_buffer *leaf,
 			  struct btrfs_key *key, int slot)
 {
-	struct btrfs_fs_info *fs_info = leaf->fs_info;
 	struct btrfs_dev_item *ditem;
-	u64 max_devid = max(BTRFS_MAX_DEVS(fs_info), BTRFS_MAX_DEVS_SYS_CHUNK);
 
 	if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
 		dev_item_err(leaf, slot,
@@ -696,12 +694,6 @@ static int check_dev_item(struct extent_buffer *leaf,
 			     key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
 		return -EUCLEAN;
 	}
-	if (key->offset > max_devid) {
-		dev_item_err(leaf, slot,
-			     "invalid devid: has=%llu expect=[0, %llu]",
-			     key->offset, max_devid);
-		return -EUCLEAN;
-	}
 	ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
 	if (btrfs_device_id(leaf, ditem) != key->offset) {
 		dev_item_err(leaf, slot,
fs/btrfs/volumes.c +9 -0
index 8b72d03738d9..56f751192a6c 100644
@@ -4901,6 +4901,15 @@ static void check_raid56_incompat_flag(struct btrfs_fs_info *info, u64 type)
 	btrfs_set_fs_incompat(info, RAID56);
 }
 
+#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
+			- sizeof(struct btrfs_chunk))		\
+			/ sizeof(struct btrfs_stripe) + 1)
+
+#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
+				- 2 * sizeof(struct btrfs_disk_key)	\
+				- 2 * sizeof(struct btrfs_chunk))	\
+				/ sizeof(struct btrfs_stripe) + 1)
+
 static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 			       u64 start, u64 type)
 {
fs/btrfs/volumes.h +0 -9
index 7f6aa1816409..789f983a98c5 100644
@@ -273,15 +273,6 @@ struct btrfs_fs_devices {
 
 #define BTRFS_BIO_INLINE_CSUM_SIZE	64
 
-#define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)	\
-			- sizeof(struct btrfs_chunk))		\
-			/ sizeof(struct btrfs_stripe) + 1)
-
-#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
-				- 2 * sizeof(struct btrfs_disk_key)	\
-				- 2 * sizeof(struct btrfs_chunk))	\
-				/ sizeof(struct btrfs_stripe) + 1)
-
 /*
  * we need the mirror number and stripe index to be passed around
  * the call chain while we are processing end_io (especially errors).
#4 lore raw

Re: [PATCH v3 2/2] btrfs: tree-checker: Fix wrong check on max devid

In reply to Qu Wenruo
Quoted 9 lines
fs/btrfs/tree-checker.c +0 -0
index ccd5706199d7..15d1aa7cef1f 100644

fs/btrfs/tree-checker.c static void dev_item_err(const struct extent_buffer *eb, int slot, · line 691

Quoted 6 lines
fs/btrfs/tree-checker.c +0 -2
@@ -686,9 +686,7 @@ static void dev_item_err(const struct extent_buffer *eb, int slot,
  static int check_dev_item(struct extent_buffer *leaf,
  			  struct btrfs_key *key, int slot)
  {
-	struct btrfs_fs_info *fs_info = leaf->fs_info;
  	struct btrfs_dev_item *ditem;
-	u64 max_devid = max(BTRFS_MAX_DEVS(fs_info), BTRFS_MAX_DEVS_SYS_CHUNK);

As I commented in v2.
I see that BTRFS_MAX_DEVS_SYS_CHUNK is not being used anywhere else after this being removed. So good to delete the define. I am bit surprised as well if I am missing?

Thanks, Anand

author
#5 lore raw

Re: [PATCH v3 2/2] btrfs: tree-checker: Fix wrong check on max devid

Quoted 14 lines · Anand Jain
Quoted 14 lines
fs/btrfs/tree-checker.c +0 -1
index ccd5706199d7..15d1aa7cef1f 100644
@@ -686,9 +686,7 @@ static void dev_item_err(const struct
 extent_buffer *eb, int slot,
   static int check_dev_item(struct extent_buffer *leaf,
                 struct btrfs_key *key, int slot)
   {
-    struct btrfs_fs_info *fs_info = leaf->fs_info;
Quoted 8 lines

fs/btrfs/tree-checker.c static void dev_item_err(const struct · line 691

Quoted 3 lines
fs/btrfs/tree-checker.c +0 -1
       struct btrfs_dev_item *ditem;
-    u64 max_devid = max(BTRFS_MAX_DEVS(fs_info),
 BTRFS_MAX_DEVS_SYS_CHUNK);

As I commented in v2.
I see that BTRFS_MAX_DEVS_SYS_CHUNK is not being used anywhere else after this being removed. So good to delete the define. I am bit surprised as well if I am missing?

Please check the first patch.

It adds back the reference to it as an early exit for btrfs_alloc_chunk().

Thanks,
Qu

Quoted 3 lines

Thanks, Anand

#6 lore raw

Re: [PATCH v3 2/2] btrfs: tree-checker: Fix wrong check on max devid

Quoted 26 lines · WenRuo Qu
Quoted 14 lines · Anand Jain
Quoted 14 lines
fs/btrfs/tree-checker.c +0 -1
index ccd5706199d7..15d1aa7cef1f 100644
@@ -686,9 +686,7 @@ static void dev_item_err(const struct
 extent_buffer *eb, int slot,
   static int check_dev_item(struct extent_buffer *leaf,
                 struct btrfs_key *key, int slot)
   {
-    struct btrfs_fs_info *fs_info = leaf->fs_info;
Quoted 8 lines

fs/btrfs/tree-checker.c static void dev_item_err(const struct · line 691

Quoted 3 lines
fs/btrfs/tree-checker.c +0 -1
       struct btrfs_dev_item *ditem;
-    u64 max_devid = max(BTRFS_MAX_DEVS(fs_info),
 BTRFS_MAX_DEVS_SYS_CHUNK);

As I commented in v2.
I see that BTRFS_MAX_DEVS_SYS_CHUNK is not being used anywhere else after this being removed. So good to delete the define. I am bit surprised as well if I am missing?

Quoted 3 lines

Please check the first patch.

It adds back the reference to it as an early exit for btrfs_alloc_chunk().

Oh. Right here in the same thread. Ok. Thanks.

Quoted 5 lines

Thanks,
Qu

Quoted 3 lines

Thanks, Anand

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

For SYSTEM chunks, despite the regular chunk item size limit, there is another limit due to system chunk array size.

The extra limit is removed in a refactor, so just add it back.

Fixes e3ecdb3fdecf ("btrfs: factor out devs_max setting in __btrfs_alloc_chunk")
Quoted 17 lines
---
 fs/btrfs/volumes.c | 1 +
 1 file changed, 1 insertion(+)
fs/btrfs/volumes.c +1 -0
index a447d3ec48d5..8b72d03738d9 100644
@@ -4966,6 +4966,7 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 	} else if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
 		max_stripe_size = SZ_32M;
 		max_chunk_size = 2 * max_stripe_size;
+		devs_max = min_t(int, devs_max, BTRFS_MAX_DEVS_SYS_CHUNK);
 	} else {
 		btrfs_err(info, "invalid chunk type 0x%llx requested",
 		       type);
Reviewed-by on v1 1/2
#8 lore raw
In reply to Qu Wenruo
Quoted 16 lines · Qu Wenruo

For SYSTEM chunks, despite the regular chunk item size limit, there is another limit due to system chunk array size.

The extra limit is removed in a refactor, so just add it back.

Fixes e3ecdb3fdecf ("btrfs: factor out devs_max setting in __btrfs_alloc_chunk")
---
  fs/btrfs/volumes.c | 1 +
  1 file changed, 1 insertion(+)
fs/btrfs/volumes.c +0 -0
index a447d3ec48d5..8b72d03738d9 100644

fs/btrfs/volumes.c static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, · line 4969

Quoted 4 lines
fs/btrfs/volumes.c +1 -0
@@ -4966,6 +4966,7 @@ static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
  	} else if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
  		max_stripe_size = SZ_32M;
  		max_chunk_size = 2 * max_stripe_size;
+		devs_max = min_t(int, devs_max, BTRFS_MAX_DEVS_SYS_CHUNK);

Thanks, Anand

Quoted 4 lines
fs/btrfs/volumes.c +0 -0
  	} else {
  		btrfs_err(info, "invalid chunk type 0x%llx requested",
  		       type);
#9 lore raw
In reply to Qu Wenruo
Quoted 7 lines · Qu Wenruo

For SYSTEM chunks, despite the regular chunk item size limit, there is another limit due to system chunk array size.

The extra limit is removed in a refactor, so just add it back.

Fixes e3ecdb3fdecf ("btrfs: factor out devs_max setting in __btrfs_alloc_chunk")

So these patches fix test btrfs/194 failure, and given the regression also worth pushing to 5.4-rc.