Skip to contentSkip to conversation

2.4.15-pre1: "bogus" message with reiserfs root and other weirdness

DISCUSSION Eric M 7 participants 13 messages Nov 21, 2001 – Nov 26, 2001

Subsystems (2)
lore
Share & tools
lore

Eric M

author
#0 lore raw

I just tried changing my root and /usr partitions from ext2 to reiserfs and noticed some weird messages on boot.

Partitions on my ide-drive are like this:

   Device Boot    Start       End    Blocks   Id  System
/dev/hda1   *         1       521   4184901    c  Win95 FAT32 (LBA)
/dev/hda2           524       654   1052257+  83  Linux
/dev/hda3           655       785   1052257+  82  Linux swap
/dev/hda4           786      5606  38724682+   5  Extended
/dev/hda5           786      1308   4200966   83  Linux
/dev/hda6          1309      5606  34523653+  83  Linux

Where hda2 is root and hda5 is /usr. The windows partition was empty at the time and i use lilo on the MBR. Lilo is version 22.1, kernel version 2.4.15-pre1. Anyway, i took backups of the root and /usr partitions with cp -a to another partition, booted from a bootdisk with reiserfs support, ran mkreiserfs and copied stuff back with cp -a. Checked that everything went back alright, ran lilo and booted.

Machine booted ok and everything seemed to be ok, but i noticed a few weird messages in boot messages right before mounting the root-partition:
FAT: bogus logical sector size 0
FAT: bogus logical sector size 0

Other reiserfs mounts didn't give this message. it doesn't seem serious but a little disturbing still.

Also, I noticed some weird interactions with the bttv-driver. I have a bt878 based tv-card.
The bttv module and all other modules it requires loaded without hickups, didn't see anything unusual in their log-messages, but as soon as i tried starting any program that uses the video-device, I got this message:
kernel: bttv: vmalloc_32(4259840) failed

As soon as i changed my root and /usr partitions back to ext2 the same way I changed to reiserfs before, these problems disappeared. Nothing else on my computer changed in between these filesystem changes. Any ideas on what might cause this? E.

#1 lore raw
Quoted 1 line

FAT: bogus logical sector size 0

It is not a problem at all.

It happens if you compile fat in the Kernel and not as module. In this case the fat-driver seems to check every partition for a fat-filesystem. When it fails it gives you this message.

So you can
1. ignore these messages or
2. compile fat as a module.

greetings

Christian

#2 lore raw
In reply to Eric M
Quoted 10 lines · Christian Bornträger
Quoted 2 lines

FAT: bogus logical sector size 0

It is not a problem at all.

It happens if you compile fat in the Kernel and not as module. In this case the fat-driver seems to check every partition for a fat-filesystem. When it fails it gives you this message.

So you can 1. ignore these messages or
2. compile fat as a module.

Quoted 10 lines

greetings

Christian
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

It is a user interface error by the FAT coder, a serious one actually, because it affects partitions that are not FAT. Would be nice if it got fixed.

Hans

#3 lore raw
In reply to Eric M
Quoted 4 lines · Eric M

Machine booted ok and everything seemed to be ok, but i noticed a few weird messages in boot messages right before mounting the root-partition:
FAT: bogus logical sector size 0
FAT: bogus logical sector size 0

When the kernel is booting, it doesn't know the filesystem type of the root fs, so it tries to mount the root device using all of the compiled-in fs drivers, in the order they are listed in fs/Makefile.in.

It appears that the fat driver doesn't even check for a magic when it starts trying to mount the filesystem, so it proceeds directly to spewing errors. You get two such messages because it is trying both msdos and vfat (which both use the fat driver underneath). You don't see this with ext2 because it tries to mount before vfat/msdos.

This is a minor bug in the fat read_super() method, because when the kernel is trying to mount the root fs, it passes the "silent" flag to the filesystem, so it should not print any messages if it is not a fat filesystem. If this wasn't the case, then you would get spew from every filesystem it tried.

Note that reiserfs is guilty of this as well, since it prints out:

can't find a reiserfs filesystem on (dev 08:03, block 64, size 4096)

So a patch for that is attached (against 2.4.13, but I didn't see any changes which would affect it).

This message is never normally seen, because currently reiserfs is at the end of the search list for filesystems. I will also propose a patch to move reiserfs earlier in the probe sequence, since it has a well-defined magic, and is more commonly used than most others. I only put it where it is to avoid context issues with my tree (which has ext3 and jbd (an experimental filesystem interface for journal devices) at the top.

Please feel free to re-submit these to Linus, I will not do so.

Cheers, Andreas
====================== reiserfs-2.4.13-silent.diff =======================

fs/reiserfs/super.c +13 -11
@@ -346,7 +356,8 @@
 		      free, SB_FREE_BLOCKS (s));
 }
 
-static int read_super_block (struct super_block * s, int size, int offset)
+static int read_super_block (struct super_block * s, int size, int offset,
+			     int silent)
 {
     struct buffer_head * bh;
     struct reiserfs_super_block * rs;
@@ -362,9 +373,10 @@
  
     rs = (struct reiserfs_super_block *)bh->b_data;
     if (!is_reiserfs_magic_string (rs)) {
-      printk ("read_super_block: "
-              "can't find a reiserfs filesystem on (dev %s, block %lu, size %d)\n",
-              kdevname(s->s_dev), bh->b_blocknr, size);
+      if (!silent)
+        printk("read_super_block: can't find a reiserfs filesystem on "
+	       "(dev %s, block %lu, size %d)\n",
+	       kdevname(s->s_dev), bh->b_blocknr, size);
       brelse (bh);
       return 1;
     }
@@ -393,11 +405,11 @@
     rs = (struct reiserfs_super_block *)bh->b_data;
     if (!is_reiserfs_magic_string (rs) ||
 	sb_blocksize(rs) != s->s_blocksize) {
-	printk ("read_super_block: "
-		"can't find a reiserfs filesystem on (dev %s, block %lu, size %d)\n",
-		kdevname(s->s_dev), bh->b_blocknr, size);
+	if (!silent)
+	  printk("read_super_block: can't find a reiserfs filesystem on "
+		 "(dev %s, block %lu, size %d)\n",
+		 kdevname(s->s_dev), bh->b_blocknr, size);
 	brelse (bh);
-	printk ("read_super_block: can't find a reiserfs filesystem on dev %s.\n", kdevname(s->s_dev));
 	return 1;
     }
     /* must check to be sure we haven't pulled an old format super out
@@ -621,7 +633,7 @@
     }
 
     if (blocks) {
-  	printk("reserfs: resize option for remount only\n");
+	printk("reiserfs: resize option for remount only\n");
 	return NULL;
     }	
 
@@ -634,9 +646,9 @@
     }
 
     /* read block (64-th 1k block), which can contain reiserfs super block */
-    if (read_super_block (s, size, REISERFS_DISK_OFFSET_IN_BYTES)) {
+    if (read_super_block (s, size, REISERFS_DISK_OFFSET_IN_BYTES, silent)) {
 	// try old format (undistributed bitmap, super block in 8-th 1k block of a device)
-	if (read_super_block (s, size, REISERFS_OLD_DISK_OFFSET_IN_BYTES)) 
+	if (read_super_block(s, size, REISERFS_OLD_DISK_OFFSET_IN_BYTES,silent))
 	    goto error;
 	else
 	    old_format = 1;

====================== reiserfs-2.4.13-order.diff =======================

fs/Makefile +1 -1
@@ -26,7 +26,10 @@
 subdir-$(CONFIG_EXT2_FS)	+= ext2
 subdir-$(CONFIG_CRAMFS)		+= cramfs
 subdir-$(CONFIG_RAMFS)		+= ramfs
+subdir-$(CONFIG_REISERFS_FS)	+= reiserfs
 subdir-$(CONFIG_CODA_FS)	+= coda
 subdir-$(CONFIG_MINIX_FS)	+= minix
 subdir-$(CONFIG_FAT_FS)	+= fat
@@ -60,7 +63,6 @@
 subdir-$(CONFIG_AUTOFS_FS)	+= autofs
 subdir-$(CONFIG_AUTOFS4_FS)	+= autofs4
 subdir-$(CONFIG_ADFS_FS)	+= adfs
-subdir-$(CONFIG_REISERFS_FS)	+= reiserfs
 subdir-$(CONFIG_DEVPTS_FS)	+= devpts
 subdir-$(CONFIG_SUN_OPENPROMFS)	+= openpromfs
 

--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/

#4 lore raw
In reply to Eric M
Quoted 9 lines
Quoted 4 lines

Machine booted ok and everything seemed to be ok, but i noticed a few weird messages in boot messages right before mounting the root-partition:
FAT: bogus logical sector size 0
FAT: bogus logical sector size 0

When the kernel is booting, it doesn't know the filesystem type of the root fs, so it tries to mount the root device using all of the compiled-in fs drivers, in the order they are listed in fs/Makefile.in.
It appears that the fat driver doesn't even check for a magic when it starts trying to mount the filesystem, so it proceeds directly to

To be complete we should also apply this patch.

fs/fat/inode.c Wed Nov 21 21:28:49 2001 renamed from fs/fat/inode.c Thu Oct 25 09:02:26 2001 +4 -2
diff -urN linux/fs/fat/inode.c linux-new/fs/fat/inode.c
@@ -609,7 +609,8 @@
                CF_LE_W(get_unaligned((unsigned short *) &b->sector_size));
        if (!logical_sector_size
            || (logical_sector_size & (logical_sector_size - 1))) {
-               printk("FAT: bogus logical sector size %d\n",
+               if (!silent)
+                   printk("FAT: bogus logical sector size %d\n",
                       logical_sector_size);
                brelse(bh);
                goto out_invalid;
@@ -618,7 +619,8 @@
        sbi->cluster_size = b->cluster_size;
        if (!sbi->cluster_size
            || (sbi->cluster_size & (sbi->cluster_size - 1))) {
-               printk("FAT: bogus cluster size %d\n", sbi->cluster_size);
+               if (!silent)
+                   printk("FAT: bogus cluster size %d\n", sbi->cluster_size);
                brelse(bh);
                goto out_invalid;
        }
#5 lore raw

In article <[email protected]> [email protected] wrote:

| > > Machine booted ok and everything seemed to be ok, but i noticed a few
| > > weird messages in boot messages right before mounting the root-partition:
| > > FAT: bogus logical sector size 0
| > > FAT: bogus logical sector size 0
| > When the kernel is booting, it doesn't know the filesystem type of the
| > root fs, so it tries to mount the root device using all of the compiled-in
| > fs drivers, in the order they are listed in fs/Makefile.in.
| > It appears that the fat driver doesn't even check for a magic when it
| > starts trying to mount the filesystem, so it proceeds directly to
| 
| To be complete we should also apply this patch.

To be totally honest I think this is the wrong way to go. Like masking the symptoms instead of treating the disease. The problem is that the FAT driver seems to not check f/s type before claiming a f/s as its own. The better solution is to put in a check before going forward with using the f/s as FAT.

Signature · 3 lines
bill davidsen <[email protected]>
  His first management concern is not solving the problem, but covering
his ass. If he lived in the middle ages he'd wear his codpiece backward.
#6 lore raw
In reply to Eric M
Quoted 30 lines · Christian Bornträger
Quoted 11 lines
Quoted 5 lines

Machine booted ok and everything seemed to be ok, but i noticed a few weird messages in boot messages right before mounting the root-partition:
FAT: bogus logical sector size 0
FAT: bogus logical sector size 0

When the kernel is booting, it doesn't know the filesystem type of the root fs, so it tries to mount the root device using all of the compiled-in fs drivers, in the order they are listed in fs/Makefile.in.
It appears that the fat driver doesn't even check for a magic when it starts trying to mount the filesystem, so it proceeds directly to

To be complete we should also apply this patch.

diff -urN linux/fs/fat/inode.c linux-new/fs/fat/inode.c

fs/fat/inode.c Wed Nov 21 21:28:49 2001 renamed from fs/fat/inode.c Thu Oct 25 09:02:26 2001 +2 -1
@@ -609,7 +609,8 @@
               CF_LE_W(get_unaligned((unsigned short *) &b->sector_size));
       if (!logical_sector_size
           || (logical_sector_size & (logical_sector_size - 1))) {
-               printk("FAT: bogus logical sector size %d\n",
+               if (!silent)
+                   printk("FAT: bogus logical sector size %d\n",
                      logical_sector_size);
               brelse(bh);
               goto out_invalid;

fs/fat/inode.c Wed Nov 21 21:28:49 2001 line 626

Quoted 9 lines
fs/fat/inode.c Wed Nov 21 21:28:49 2001 renamed from fs/fat/inode.c Thu Oct 25 09:02:26 2001 +2 -1
@@ -618,7 +619,8 @@
       sbi->cluster_size = b->cluster_size;
       if (!sbi->cluster_size
           || (sbi->cluster_size & (sbi->cluster_size - 1))) {
-               printk("FAT: bogus cluster size %d\n", sbi->cluster_size);
+               if (!silent)
+                   printk("FAT: bogus cluster size %d\n", sbi->cluster_size);
               brelse(bh);
               goto out_invalid;
       }

How about putting fat last in the list, and having it say something more like "FAT: bogus cluster size, perhaps XXX is not a FAT partition"

maintainer
#7 lore raw
In reply to Eric M
Quoted 13 lines · [email protected] (bill davidsen) writes:

In article <[email protected]> [email protected] wrote:

| > > Machine booted ok and everything seemed to be ok, but i noticed a few
| > > weird messages in boot messages right before mounting the root-partition:
| > > FAT: bogus logical sector size 0
| > > FAT: bogus logical sector size 0
| > When the kernel is booting, it doesn't know the filesystem type of the
| > root fs, so it tries to mount the root device using all of the compiled-in
| > fs drivers, in the order they are listed in fs/Makefile.in.
| > It appears that the fat driver doesn't even check for a magic when it
| > starts trying to mount the filesystem, so it proceeds directly to
| 
| To be complete we should also apply this patch.
Quoted 5 lines

To be totally honest I think this is the wrong way to go. Like masking the symptoms instead of treating the disease. The problem is that the FAT driver seems to not check f/s type before claiming a f/s as its own. The better solution is to put in a check before going forward with using the f/s as FAT.

FAT doesn't have the field like the magic number, AFAIK.

Signature · 1 line
OGAWA Hirofumi <[email protected]>
#8 lore raw
Quoted 15 lines · OGAWA Hirofumi
Quoted 13 lines · [email protected] (bill davidsen) writes:

In article <[email protected]> [email protected] wrote:

| > > Machine booted ok and everything seemed to be ok, but i noticed a few
| > > weird messages in boot messages right before mounting the root-partition:
| > > FAT: bogus logical sector size 0
| > > FAT: bogus logical sector size 0
| > When the kernel is booting, it doesn't know the filesystem type of the
| > root fs, so it tries to mount the root device using all of the compiled-in
| > fs drivers, in the order they are listed in fs/Makefile.in.
| > It appears that the fat driver doesn't even check for a magic when it
| > starts trying to mount the filesystem, so it proceeds directly to
| 
| To be complete we should also apply this patch.
Quoted 7 lines
Quoted 5 lines

To be totally honest I think this is the wrong way to go. Like masking the symptoms instead of treating the disease. The problem is that the FAT driver seems to not check f/s type before claiming a f/s as its own. The better solution is to put in a check before going forward with using the f/s as FAT.

FAT doesn't have the field like the magic number, AFAIK.

Actually the original poster used "magic" and I didn't quibble, but I believe you're right. I just noted that it is better to check to see if the filesystem is FAT before reporting the error message. Obviously this can be done, since you don't get the message with FAT as a module, I just haven't looked to see if the check is in the module or the kernel before even loading the FAT module.

Thanks for noting that, in any case.

Signature · 3 lines
bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
maintainer
#9 lore raw
Quoted 10 lines · Bill Davidsen
Quoted 1 line · OGAWA Hirofumi

[email protected] (bill davidsen) writes:

Actually the original poster used "magic" and I didn't quibble, but I believe you're right. I just noted that it is better to check to see if the filesystem is FAT before reporting the error message. Obviously this can be done, since you don't get the message with FAT as a module, I just haven't looked to see if the check is in the module or the kernel before even loading the FAT module.

Yes. You can't use the module before you mounted the root filesystem. This is a reason for not seeing an error message.

Signature · 1 line
OGAWA Hirofumi <[email protected]>
#10 lore raw
Quoted 13 lines · OGAWA Hirofumi
Quoted 10 lines · Bill Davidsen
Quoted 1 line · OGAWA Hirofumi

[email protected] (bill davidsen) writes:

Actually the original poster used "magic" and I didn't quibble, but I believe you're right. I just noted that it is better to check to see if the filesystem is FAT before reporting the error message. Obviously this can be done, since you don't get the message with FAT as a module, I just haven't looked to see if the check is in the module or the kernel before even loading the FAT module.

Quoted 2 lines

Yes. You can't use the module before you mounted the root filesystem. This is a reason for not seeing an error message.

What? Of course you can! That's what initd files are for ;-)

Signature · 3 lines
bill davidsen <[email protected]>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
#11 lore raw
In reply to Bill Davidsen

Modular or monolithic, the bogus messages are the same.

Quoted 3 lines

Yes. You can't use the module before you mounted the root filesystem. This is a reason for not seeing an error message.