Skip to contentSkip to conversation

AOSP: erofs-utils: pass a parameter to write tail end in block list

PATCH In review Yue Hu 2 participants 10 messages Aug 15, 2023 Library code

lore
Share & tools
lore

Yue Hu

author
#0 lore raw

PATCH v2 1/2 AOSP: erofs-utils: pass a parameter to write tail end in block list

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

---
v2: change commit message a bit

 include/erofs/block_list.h | 4 ++--
 lib/block_list.c           | 5 ++---
 lib/inode.c                | 9 +++++++--
 3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +2 -2
index 78fab44..e0dced8 100644
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
 void erofs_droid_blocklist_write(struct erofs_inode *inode,
 				 erofs_blk_t blk_start, erofs_blk_t nblocks);
 void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);
 void erofs_droid_blocklist_write_extent(struct erofs_inode *inode,
 					erofs_blk_t blk_start, erofs_blk_t nblocks,
 					bool first_extent, bool last_extent);
@@ -28,7 +28,7 @@ static inline void erofs_droid_blocklist_write(struct erofs_inode *inode,
 				 erofs_blk_t blk_start, erofs_blk_t nblocks) {}
 static inline void
 erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr) {}
+				     erofs_blk_t blkaddr, bool first_block) {}
 static inline void
 erofs_droid_blocklist_write_extent(struct erofs_inode *inode,
 				   erofs_blk_t blk_start, erofs_blk_t nblocks,
lib/block_list.c +2 -3
index 896fb01..a1c719d 100644
@@ -85,7 +85,7 @@ void erofs_droid_blocklist_write(struct erofs_inode *inode,
 }
 
 void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr)
+					  erofs_blk_t blkaddr, bool first_block)
 {
 	if (!block_list_fp || !cfg.mount_point)
 		return;
@@ -94,8 +94,7 @@ void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
 	if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
 		return;
 
-	/* XXX: another hack, which means it has been outputed before */
-	if (erofs_blknr(inode->i_size)) {
+	if (!first_block) {
 		if (blkaddr == NULL_ADDR)
 			fprintf(block_list_fp, "\n");
 		else
lib/inode.c +7 -2
index c4d1476..7cd3e69 100644
@@ -734,12 +734,15 @@ static int erofs_write_tail_end(struct erofs_inode *inode)
 {
 	struct erofs_sb_info *sbi = inode->sbi;
 	struct erofs_buffer_head *bh, *ibh;
+	bool first_block;
 
 	bh = inode->bh_data;
 
 	if (!inode->idata_size)
 		goto out;
 
+	first_block = !erofs_blknr(sbi, inode->i_size);
+
 	/* have enough room to inline data */
 	if (inode->bh_inline) {
 		ibh = inode->bh_inline;
@@ -747,7 +750,8 @@ static int erofs_write_tail_end(struct erofs_inode *inode)
 		ibh->fsprivate = erofs_igrab(inode);
 		ibh->op = &erofs_write_inline_bhops;
 
-		erofs_droid_blocklist_write_tail_end(inode, NULL_ADDR);
+		erofs_droid_blocklist_write_tail_end(inode, NULL_ADDR,
+						     first_block);
 	} else {
 		int ret;
 		erofs_off_t pos, zero_pos;
@@ -801,7 +805,8 @@ static int erofs_write_tail_end(struct erofs_inode *inode)
 		free(inode->idata);
 		inode->idata = NULL;
 
-		erofs_droid_blocklist_write_tail_end(inode, erofs_blknr(sbi, pos));
+		erofs_droid_blocklist_write_tail_end(inode, erofs_blknr(sbi, pos),
+						     first_block);
 	}
 out:
 	/* now bh_data can drop directly */
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH 2/2 AOSP: erofs-utils: add a variable to validate tail block map in block list

From: Yue Hu <[email protected]>

We write the block map directly before the tail, as this applies only to regular files. However, the tail block map does not follow the rule.

Therefore, introduce a variable to track whether the tail block is safe to write before its map if it exists rather than separately check the file type when mapping.

---
 lib/block_list.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)
lib/block_list.c +9 -7
index a1c719d..e59a8d7 100644
@@ -12,6 +12,7 @@
 #include "erofs/print.h"
 
 static FILE *block_list_fp;
+static bool regular_tail = false;
 
 int erofs_droid_blocklist_fopen(void)
 {
@@ -77,22 +78,23 @@ void erofs_droid_blocklist_write_extent(struct erofs_inode *inode,
 void erofs_droid_blocklist_write(struct erofs_inode *inode,
 				 erofs_blk_t blk_start, erofs_blk_t nblocks)
 {
-	if (!block_list_fp || !cfg.mount_point || !nblocks)
+	if (!block_list_fp || !cfg.mount_point)
 		return;
 
-	blocklist_write(inode->i_srcpath, blk_start, nblocks,
-			true, !inode->idata_size);
+	if (nblocks)
+		blocklist_write(inode->i_srcpath, blk_start, nblocks, true,
+				!inode->idata_size);
+	if (inode->idata_size)
+		regular_tail = true;
 }
 
 void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
 					  erofs_blk_t blkaddr, bool first_block)
 {
-	if (!block_list_fp || !cfg.mount_point)
+	if (!block_list_fp || !cfg.mount_point || !regular_tail)
 		return;
 
-	/* XXX: a bit hacky.. may need a better approach */
-	if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
-		return;
+	regular_tail = false;
 
 	if (!first_block) {
 		if (blkaddr == NULL_ADDR)
Signature · 1 line
2.17.1
#2 lore raw
In reply to Yue Hu
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

Quoted 15 lines

---
v2: change commit message a bit

  include/erofs/block_list.h | 4 ++--
  lib/block_list.c           | 5 ++---
  lib/inode.c                | 9 +++++++--
  3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +0 -0
index 78fab44..e0dced8 100644

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 5 lines
include/erofs/block_list.h +1 -1
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
  void erofs_droid_blocklist_write(struct erofs_inode *inode,
  				 erofs_blk_t blk_start, erofs_blk_t nblocks);
  void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);

I still have no idea why we need this, could you describe the Android block map details for discussion?

Thanks,
Gao Xiang

author
#3 lore raw
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 11 lines
Quoted 11 lines

---
v2: change commit message a bit

  include/erofs/block_list.h | 4 ++--
  lib/block_list.c           | 5 ++---
  lib/inode.c                | 9 +++++++--
  3 files changed, 11 insertions(+), 7 deletions(-)
Quoted 13 lines

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 10 lines
include/erofs/block_list.h +1 -1
index 78fab44..e0dced8 100644
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
  void erofs_droid_blocklist_write(struct erofs_inode *inode,
  				 erofs_blk_t blk_start, erofs_blk_t nblocks);
  void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);  

I still have no idea why we need this, could you describe the Android block map details for discussion?

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

or

/file_path tail_block_address //only one block

Quoted 3 lines

Thanks,
Gao Xiang

#4 lore raw
Quoted 41 lines · Yue Hu
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 11 lines
Quoted 11 lines

---
v2: change commit message a bit

   include/erofs/block_list.h | 4 ++--
   lib/block_list.c           | 5 ++---
   lib/inode.c                | 9 +++++++--
   3 files changed, 11 insertions(+), 7 deletions(-)
Quoted 13 lines

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 10 lines
include/erofs/block_list.h +1 -1
index 78fab44..e0dced8 100644
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
   void erofs_droid_blocklist_write(struct erofs_inode *inode,
   				 erofs_blk_t blk_start, erofs_blk_t nblocks);
   void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);

I still have no idea why we need this, could you describe the Android block map details for discussion?

Quoted 6 lines

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

why tail_block_address needs a seperate field?

Quoted 4 lines

or

/file_path tail_block_address //only one block

how about recording the previous inode pointer as an internal global variable? if it's not the same inode, trigger "\n" immediately.

and generate files as:
" block1_address-blockn_address" for many blocks; " block1_address" for a single block?

Thanks,
Gao Xiang

author
#5 lore raw
Quoted 42 lines · On Tue, 15 Aug 2023 13:36:46 +0800 Gao Xiang <[email protected]>…
Quoted 41 lines · Yue Hu
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 24 lines
Quoted 16 lines

---
v2: change commit message a bit

   include/erofs/block_list.h | 4 ++--
   lib/block_list.c           | 5 ++---
   lib/inode.c                | 9 +++++++--
   3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +0 -0
index 78fab44..e0dced8 100644

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 5 lines
include/erofs/block_list.h +1 -1
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
   void erofs_droid_blocklist_write(struct erofs_inode *inode,
   				 erofs_blk_t blk_start, erofs_blk_t nblocks);
   void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);  

I still have no idea why we need this, could you describe the Android block map details for discussion?

Quoted 8 lines
Quoted 6 lines

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

why tail_block_address needs a seperate field?

Well, erofs_write_tail_end() is a separate logic and i think appending this block is simple enough since i don't need to consider whether the block address is contiguous with previous one.

And i think Android block map can handle this since i have saw below in ext4:

/system/.../libclang_rt.ubsan_standalone-arm-android.so 51276-51309 0 51310-51403

Quoted 4 lines
Quoted 4 lines

or

Quoted 8 lines
Quoted 1 line

/file_path tail_block_address //only one block

how about recording the previous inode pointer as an internal global variable? if it's not the same inode, trigger "\n" immediately.

and generate files as:
" block1_address-blockn_address" for many blocks; " block1_address" for a single block?

Let me check.

Quoted 3 lines

Thanks,
Gao Xiang

#6 lore raw
Quoted 54 lines · Yue Hu
Quoted 42 lines · On Tue, 15 Aug 2023 13:36:46 +0800 Gao Xiang <[email protected]>…
Quoted 41 lines · Yue Hu
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 24 lines
Quoted 16 lines

---
v2: change commit message a bit

    include/erofs/block_list.h | 4 ++--
    lib/block_list.c           | 5 ++---
    lib/inode.c                | 9 +++++++--
    3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +0 -0
index 78fab44..e0dced8 100644

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 5 lines
include/erofs/block_list.h +1 -1
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
    void erofs_droid_blocklist_write(struct erofs_inode *inode,
    				 erofs_blk_t blk_start, erofs_blk_t nblocks);
    void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);

I still have no idea why we need this, could you describe the Android block map details for discussion?

Quoted 8 lines
Quoted 6 lines

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

why tail_block_address needs a seperate field?

Quoted 7 lines

Well, erofs_write_tail_end() is a separate logic and i think appending this block is simple enough since i don't need to consider whether the block address is contiguous with previous one.

And i think Android block map can handle this since i have saw below in ext4:

/system/.../libclang_rt.ubsan_standalone-arm-android.so 51276-51309 0 51310-51403

What's the meaning of 0 here?

Thanks,
Gao Xiang

author
#7 lore raw
Quoted 59 lines · On Tue, 15 Aug 2023 14:18:08 +0800 Gao Xiang <[email protected]>…
Quoted 58 lines · Yue Hu
Quoted 42 lines · On Tue, 15 Aug 2023 13:36:46 +0800 Gao Xiang <[email protected]>…
Quoted 41 lines · Yue Hu
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 24 lines
Quoted 16 lines

---
v2: change commit message a bit

    include/erofs/block_list.h | 4 ++--
    lib/block_list.c           | 5 ++---
    lib/inode.c                | 9 +++++++--
    3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +0 -0
index 78fab44..e0dced8 100644

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 5 lines
include/erofs/block_list.h +1 -1
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
    void erofs_droid_blocklist_write(struct erofs_inode *inode,
    				 erofs_blk_t blk_start, erofs_blk_t nblocks);
    void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);  

I still have no idea why we need this, could you describe the Android block map details for discussion?

Quoted 8 lines
Quoted 6 lines

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

why tail_block_address needs a seperate field?

Well, erofs_write_tail_end() is a separate logic and i think appending this block is simple enough since i don't need to consider whether the block address is contiguous with previous one.

Quoted 5 lines
Quoted 3 lines

And i think Android block map can handle this since i have saw below in ext4:

/system/.../libclang_rt.ubsan_standalone-arm-android.so 51276-51309 0 51310-51403

What's the meaning of 0 here?

I have not check that yet, maybe i need to find time...

And check this: "/.../libxxx.so 87378-88630 0 0 0 88631 0 0 88632-88637 0 0 0 88638-88663"

so, i guess it looks like null block.

Quoted 4 lines

Thanks,
Gao Xiang

#8 lore raw
Quoted 66 lines · Yue Hu
Quoted 63 lines · On Tue, 15 Aug 2023 14:18:08 +0800 Gao Xiang <[email protected]>…
Quoted 61 lines · Yue Hu
Quoted 42 lines · On Tue, 15 Aug 2023 13:36:46 +0800 Gao Xiang <[email protected]>…
Quoted 41 lines · Yue Hu
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 24 lines
Quoted 16 lines

---
v2: change commit message a bit

     include/erofs/block_list.h | 4 ++--
     lib/block_list.c           | 5 ++---
     lib/inode.c                | 9 +++++++--
     3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +0 -0
index 78fab44..e0dced8 100644

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 5 lines
include/erofs/block_list.h +1 -1
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
     void erofs_droid_blocklist_write(struct erofs_inode *inode,
     				 erofs_blk_t blk_start, erofs_blk_t nblocks);
     void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);

I still have no idea why we need this, could you describe the Android block map details for discussion?

Quoted 8 lines
Quoted 6 lines

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

why tail_block_address needs a seperate field?

Well, erofs_write_tail_end() is a separate logic and i think appending this block is simple enough since i don't need to consider whether the block address is contiguous with previous one.

And i think Android block map can handle this since i have saw below in ext4:

/system/.../libclang_rt.ubsan_standalone-arm-android.so 51276-51309 0 51310-51403

Quoted 7 lines
Quoted 1 line

What's the meaning of 0 here?

I have not check that yet, maybe i need to find time...

And check this: "/.../libxxx.so 87378-88630 0 0 0 88631 0 0 88632-88637 0 0 0 88638-88663"

so, i guess it looks like null block.

So here erofs_droid_blocklist_write_tail_end is not actually needed in principle? 0 is just used for sparse block?

Thanks,
Gao Xiang

author
#9 lore raw
Quoted 69 lines · On Tue, 15 Aug 2023 14:44:23 +0800 Gao Xiang <[email protected]>…
Quoted 68 lines · Yue Hu
Quoted 64 lines · On Tue, 15 Aug 2023 14:18:08 +0800 Gao Xiang <[email protected]>…
Quoted 54 lines · Yue Hu
Quoted 42 lines · On Tue, 15 Aug 2023 13:36:46 +0800 Gao Xiang <[email protected]>…
Quoted 41 lines · Yue Hu
Quoted 10 lines · On Tue, 15 Aug 2023 12:59:56 +0800 Gao Xiang <[email protected]>…
Quoted 7 lines · Yue Hu

From: Yue Hu <[email protected]>

We can determine whether the tail block is the first one or not during the writing process. Therefore, instead of internally checking the block number for the tail block map, just simply pass the flag.

Also, add the missing sbi argument to macro erofs_blknr.

Could you submit a patch to fix this issue first?

ok, will do that.

Quoted 24 lines
Quoted 16 lines

---
v2: change commit message a bit

     include/erofs/block_list.h | 4 ++--
     lib/block_list.c           | 5 ++---
     lib/inode.c                | 9 +++++++--
     3 files changed, 11 insertions(+), 7 deletions(-)
include/erofs/block_list.h +0 -0
index 78fab44..e0dced8 100644

include/erofs/block_list.h void erofs_droid_blocklist_fclose(void); · line 22

Quoted 5 lines
include/erofs/block_list.h +1 -1
@@ -19,7 +19,7 @@ void erofs_droid_blocklist_fclose(void);
     void erofs_droid_blocklist_write(struct erofs_inode *inode,
     				 erofs_blk_t blk_start, erofs_blk_t nblocks);
     void erofs_droid_blocklist_write_tail_end(struct erofs_inode *inode,
-					  erofs_blk_t blkaddr);
+					  erofs_blk_t blkaddr, bool first_block);  

I still have no idea why we need this, could you describe the Android block map details for discussion?

Quoted 8 lines
Quoted 6 lines

Android block map is just adding file blocks to a range.

So, the tail block should be needed in this range as well.
I think one simple way is just appending the tail block address in it as below:

/file_path block1_address-blockn_address tail_block_address

why tail_block_address needs a seperate field?

Quoted 7 lines

Well, erofs_write_tail_end() is a separate logic and i think appending this block is simple enough since i don't need to consider whether the block address is contiguous with previous one.

And i think Android block map can handle this since i have saw below in ext4:

/system/.../libclang_rt.ubsan_standalone-arm-android.so 51276-51309 0 51310-51403

What's the meaning of 0 here?

Quoted 8 lines
Quoted 5 lines

I have not check that yet, maybe i need to find time...

And check this: "/.../libxxx.so 87378-88630 0 0 0 88631 0 0 88632-88637 0 0 0 88638-88663"

so, i guess it looks like null block.

So here erofs_droid_blocklist_write_tail_end is not actually needed in principle? 0 is just used for sparse block?

We need to confirm this.

Quoted 3 lines

Thanks,
Gao Xiang