Skip to contentSkip to conversation

isofs: Fix lseek() to position beyond 4 GB

PATCH Superseded Jan Andres 3 participants 5 messages Aug 3, 2010 – Sep 30, 2010

Subsystems (2)
lore
Share & tools
lore

Jan Andres

author
#0 lore raw

PATCH 1/1 isofs: Fix lseek() to position beyond 4 GB

isofs supports files larger than 4 GB by using multi-extent files. However an lseek() to a position beyond 4 GB in such a file will fail with EINVAL, because s_maxbytes in the isofs superblock is initialized to 2^32-1, and generic_file_llseek() checks against that value.

I therefore suggest increasing the value of s_maxbytes to have full support for large files in isofs. With multi-extent files, file size is only limited by the maximum size of the file system (8 TB), so this seems a reasonable value for s_maxbytes.

fs/isofs/inode.c renamed from fs/isofs/inode.c.orig +6 -1
@@ -722,7 +722,12 @@ root_found:
 	}
 
 	s->s_magic = ISOFS_SUPER_MAGIC;
-	s->s_maxbytes = 0xffffffff; /* We can handle files up to 4 GB */
+
+	/*
+	 * With multi-extent files, file size is only limited by the maximum
+	 * size of a file system, which is 8 TB.
+	 */
+	s->s_maxbytes = 0x80000000000;
 
 	/*
 	 * The CDROM is read-only, has no nodes (devices) on it, and since
#1 lore raw

fs/isofs/inode.c root_found: · line 730

Quoted 8 lines · Jan Andres

isofs supports files larger than 4 GB by using multi-extent files. However an lseek() to a position beyond 4 GB in such a file will fail with EINVAL, because s_maxbytes in the isofs superblock is initialized to 2^32-1, and generic_file_llseek() checks against that value.

fs/isofs/inode.c renamed from fs/isofs/inode.c.orig +1 -0
+	s->s_maxbytes = 0x80000000000;

0x80000000000LL;

author
#2 lore raw

PATCH v2 1/1 isofs: Fix lseek() to position beyond 4 GB

In reply to Jan Andres

isofs supports files larger than 4 GB by using multi-extent files. However an lseek() to a position beyond 4 GB in such a file will fail with EINVAL, because s_maxbytes in the isofs superblock is initialized to 2^32-1, and generic_file_llseek() checks against that value.

I therefore suggest increasing the value of s_maxbytes to have full support for large files in isofs. With multi-extent files, file size is only limited by the maximum size of the file system (8 TB), so this seems a reasonable value for s_maxbytes.

fs/isofs/inode.c renamed from fs/isofs/inode.c.orig +6 -1
@@ -722,7 +722,12 @@ root_found:
 	}
 
 	s->s_magic = ISOFS_SUPER_MAGIC;
-	s->s_maxbytes = 0xffffffff; /* We can handle files up to 4 GB */
+
+	/*
+	 * With multi-extent files, file size is only limited by the maximum
+	 * size of a file system, which is 8 TB.
+	 */
+	s->s_maxbytes = 0x80000000000LL;
 
 	/*
 	 * The CDROM is read-only, has no nodes (devices) on it, and since
maintainer
#3 lore raw
Quoted 10 lines · Jan Andres

isofs supports files larger than 4 GB by using multi-extent files. However an lseek() to a position beyond 4 GB in such a file will fail with EINVAL, because s_maxbytes in the isofs superblock is initialized to 2^32-1, and generic_file_llseek() checks against that value.

I therefore suggest increasing the value of s_maxbytes to have full support for large files in isofs. With multi-extent files, file size is only limited by the maximum size of the file system (8 TB), so this seems a reasonable value for s_maxbytes.

OK, I had a look into the relevant code and it seems that the isofs
code should properly handle files upto 4TB and with a small fix (attached) upto 8TB as you claim. So I'll take your patch and merge it with Linus.

Honza

Signature · 3 lines
Jan Kara <[email protected]>
SUSE Labs, CR
--
Quoted 1 line

From 258d9d44217700e2a7b513b9a029ad28daeca7b0 Mon Sep 17 00:00:00 2001

From: Jan Kara <[email protected]>
Date: Fri, 1 Oct 2010 00:05:33 +0200
Subject: [PATCH] isofs: Fix isofs_get_blocks for 8TB files

Currently isofs_get_blocks() was limited to handle only 4TB files on 32-bit architectures because of unnecessary use of iblock variable which was signed long. Just remove the variable. The error messages that were using this variable should have rather used b_off anyway because that is the block we are currently mapping.

---
 fs/isofs/inode.c |   18 ++++++++----------
 1 files changed, 8 insertions(+), 10 deletions(-)
fs/isofs/inode.c +8 -10
index d8a68e7..a2de39d 100644
@@ -962,26 +962,23 @@ static int isofs_statfs (struct dentry *dentry, struct kstatfs *buf)
  * or getblk() if they are not.  Returns the number of blocks inserted
  * (-ve == error.)
  */
-int isofs_get_blocks(struct inode *inode, sector_t iblock_s,
+int isofs_get_blocks(struct inode *inode, sector_t iblock,
 		     struct buffer_head **bh, unsigned long nblocks)
 {
-	unsigned long b_off;
+	unsigned long b_off = iblock;
 	unsigned offset, sect_size;
 	unsigned int firstext;
 	unsigned long nextblk, nextoff;
-	long iblock = (long)iblock_s;
 	int section, rv, error;
 	struct iso_inode_info *ei = ISOFS_I(inode);
 
 	error = -EIO;
 	rv = 0;
-	if (iblock < 0 || iblock != iblock_s) {
+	if (iblock != b_off) {
 		printk(KERN_DEBUG "%s: block number too large\n", __func__);
 		goto abort;
 	}
 
-	b_off = iblock;
-
 	offset = 0;
 	firstext = ei->i_first_extent;
 	sect_size = ei->i_section_size >> ISOFS_BUFFER_BITS(inode);
@@ -998,8 +995,9 @@ int isofs_get_blocks(struct inode *inode, sector_t iblock_s,
 		 * I/O errors.
 		 */
 		if (b_off > ((inode->i_size + PAGE_CACHE_SIZE - 1) >> ISOFS_BUFFER_BITS(inode))) {
-			printk(KERN_DEBUG "%s: block >= EOF (%ld, %ld)\n",
-				__func__, iblock, (unsigned long) inode->i_size);
+			printk(KERN_DEBUG "%s: block >= EOF (%lu, %llu)\n",
+				__func__, b_off,
+				(unsigned long long)inode->i_size);
 			goto abort;
 		}
 
@@ -1025,9 +1023,9 @@ int isofs_get_blocks(struct inode *inode, sector_t iblock_s,
 			if (++section > 100) {
 				printk(KERN_DEBUG "%s: More than 100 file sections ?!?"
 					" aborting...\n", __func__);
-				printk(KERN_DEBUG "%s: block=%ld firstext=%u sect_size=%u "
+				printk(KERN_DEBUG "%s: block=%lu firstext=%u sect_size=%u "
 					"nextblk=%lu nextoff=%lu\n", __func__,
-					iblock, firstext, (unsigned) sect_size,
+					b_off, firstext, (unsigned) sect_size,
 					nextblk, nextoff);
 				goto abort;
 			}
Signature · 1 line
1.6.4.2
maintainer
#4 lore raw
Quoted 7 lines · Jan Kara
Quoted 6 lines · Jan Andres

isofs supports files larger than 4 GB by using multi-extent files. However an lseek() to a position beyond 4 GB in such a file will fail with EINVAL, because s_maxbytes in the isofs superblock is initialized to 2^32-1, and generic_file_llseek() checks against that value.

Quoted 7 lines
Quoted 4 lines

I therefore suggest increasing the value of s_maxbytes to have full support for large files in isofs. With multi-extent files, file size is only limited by the maximum size of the file system (8 TB), so this seems a reasonable value for s_maxbytes.

OK, I had a look into the relevant code and it seems that the isofs
code should properly handle files upto 4TB and with a small fix (attached) upto 8TB as you claim. So I'll take your patch and merge it with Linus.

Oops, I now noticed that Al has already merged your patch. OK, so I'll
just merge my fix separately.

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR