Re: [patch set 4.19.6] BFS updates

[Tigran Aivazian <aivazian.tigran@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 756 bytes --]

Re-sending the patches with the "Cc: stable@vger.kernel.org" included
in the sign-off area, as per Option 1 of the rules Greg referred me
to. I hope that now I have done everything by the rules.

On Sun, 2 Dec 2018 at 21:42, Tigran Aivazian <aivazian.tigran@gmail.com> wrote:
>
> just wanted to add: although the subject says "4.19.6" the patches
> apply perfectly to the top of "torvalds/linux" tree from github.
> On Sun, 2 Dec 2018 at 21:01, Tigran Aivazian <aivazian.tigran@gmail.com> wrote:
> >
> > Hi Linus,
> >
> > I attached two incremental patches for BFS:
> >
> > 1. Make inode bitmap allocation static (applies on top of 4.19.6)
> > 2. Strengthen the superblock sanity checking code (applies on top of 1. above)
> >
> > Kind regards,
> > Tigran

[-- Attachment #2: bfs-4.19.6-1-of-2-static-inode-bitmap.patch --]
[-- Type: text/x-patch, Size: 3506 bytes --]

From: Tigran Aivazian <aivazian.tigran@gmail.com>
Subject: [PATCH 4.19.6 1/2] BFS updates

Make in-core inode bitmap static part of superblock info structure an.
print a warning when mounting a BFS filesystem created with "-N 512"
option as only 510 files can be created in the root directory.

Signed-off-by: Tigran Aivazian <aivazian.tigran@gmail.com>
Cc: stable@vger.kernel.org
---

 bfs.h   |    9 ++++++++-
 inode.c |   27 +++++++++++----------------
 2 files changed, 19 insertions(+), 17 deletions(-)

--- fs/bfs/bfs.h.0	2018-12-02 20:33:02.252710291 +0000
+++ fs/bfs/bfs.h	2018-12-02 20:34:34.041246489 +0000
@@ -8,6 +8,13 @@
 
 #include <linux/bfs_fs.h>
 
+/* In theory BFS supports up to 512 inodes, numbered from 2 (for /) up to 513 inclusive.
+   In actual fact, attempting to create the 512th inode (i.e. inode No. 513 or file No. 511)
+   will fail with ENOSPC in bfs_add_entry(): the root directory cannot contain so many entries, counting '..'.
+   So, mkfs.bfs(8) should really limit its -N option to 511 and not 512. For now, we just print a warning
+   if a filesystem is mounted with such "impossible to fill up" number of inodes */
+#define BFS_MAX_LASTI	513
+
 /*
  * BFS file system in-core superblock info
  */
@@ -17,7 +24,7 @@
 	unsigned long si_freei;
 	unsigned long si_lf_eblk;
 	unsigned long si_lasti;
-	unsigned long *si_imap;
+	DECLARE_BITMAP(si_imap, BFS_MAX_LASTI+1);
 	struct mutex bfs_lock;
 };
 
--- fs/bfs/inode.c.0	2018-12-02 20:34:03.211740877 +0000
+++ fs/bfs/inode.c	2018-12-02 20:36:54.508963813 +0000
@@ -214,7 +214,6 @@
 		return;
 
 	mutex_destroy(&info->bfs_lock);
-	kfree(info->si_imap);
 	kfree(info);
 	s->s_fs_info = NULL;
 }
@@ -322,7 +321,7 @@
 	struct buffer_head *bh, *sbh;
 	struct bfs_super_block *bfs_sb;
 	struct inode *inode;
-	unsigned i, imap_len;
+	unsigned i;
 	struct bfs_sb_info *info;
 	int ret = -EINVAL;
 	unsigned long i_sblock, i_eblock, i_eoff, s_size;
@@ -356,13 +355,11 @@
 		goto out1;
 	}
 
-	info->si_lasti = (le32_to_cpu(bfs_sb->s_start) - BFS_BSIZE) /
-					sizeof(struct bfs_inode)
-					+ BFS_ROOT_INO - 1;
-	imap_len = (info->si_lasti / 8) + 1;
-	info->si_imap = kzalloc(imap_len, GFP_KERNEL | __GFP_NOWARN);
-	if (!info->si_imap) {
-		printf("Cannot allocate %u bytes\n", imap_len);
+	info->si_lasti = (le32_to_cpu(bfs_sb->s_start) - BFS_BSIZE) / sizeof(struct bfs_inode) + BFS_ROOT_INO - 1;
+	if (info->si_lasti == BFS_MAX_LASTI)
+		printf("WARNING: filesystem %s was created with 512 inodes, the real maximum is 511, mounting anyway\n", s->s_id);
+	else if (info->si_lasti > BFS_MAX_LASTI) {
+		printf("Impossible last inode number %lu > %d on %s\n", info->si_lasti, BFS_MAX_LASTI, s->s_id);
 		goto out1;
 	}
 	for (i = 0; i < BFS_ROOT_INO; i++)
@@ -372,12 +369,12 @@
 	inode = bfs_iget(s, BFS_ROOT_INO);
 	if (IS_ERR(inode)) {
 		ret = PTR_ERR(inode);
-		goto out2;
+		goto out1;
 	}
 	s->s_root = d_make_root(inode);
 	if (!s->s_root) {
 		ret = -ENOMEM;
-		goto out2;
+		goto out1;
 	}
 
 	info->si_blocks = (le32_to_cpu(bfs_sb->s_end) + 1) >> BFS_BSIZE_BITS;
@@ -391,7 +388,7 @@
 	if (!bh) {
 		printf("Last block not available: %lu\n", info->si_blocks - 1);
 		ret = -EIO;
-		goto out3;
+		goto out2;
 	}
 	brelse(bh);
 
@@ -429,7 +426,7 @@
 
 			brelse(bh);
 			ret = -EIO;
-			goto out3;
+			goto out2;
 		}
 
 		if (!di->i_ino) {
@@ -448,11 +445,9 @@
 	bfs_dump_imap("read_super", s);
 	return 0;
 
-out3:
+out2:
 	dput(s->s_root);
 	s->s_root = NULL;
-out2:
-	kfree(info->si_imap);
 out1:
 	brelse(sbh);
 out:

[-- Attachment #3: bfs-4.19.6-2-of-2-strengthen-sanity-checking.patch --]
[-- Type: text/x-patch, Size: 771 bytes --]

From: Tigran Aivazian <aivazian.tigran@gmail.com>
Subject: [PATCH 4.19.6 2/2] BFS updates

Strengthen the superblock sanity checking (supersedes the code
that went into 4.19.6).

Signed-off-by: Tigran Aivazian <aivazian.tigran@gmail.com>
Cc: stable@vger.kernel.org
---

 inode.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- fs/bfs/inode.c.0	2018-12-02 20:49:00.525461354 +0000
+++ fs/bfs/inode.c	2018-12-02 20:49:51.054686779 +0000
@@ -350,7 +350,7 @@
 	s->s_magic = BFS_MAGIC;
 
 	if (le32_to_cpu(bfs_sb->s_start) > le32_to_cpu(bfs_sb->s_end) ||
-	    le32_to_cpu(bfs_sb->s_start) < BFS_BSIZE) {
+	    le32_to_cpu(bfs_sb->s_start) < sizeof(struct bfs_super_block) + sizeof(struct bfs_dirent)) {
 		printf("Superblock is corrupted\n");
 		goto out1;
 	}