[U-Boot] genext2fs fix for supporting filesystems generated by genext2fs
I found that the latest U-Boot cannot handle filesystems generated by genext2fs. Attempts to list such a filesystem results in divide by zero.
Using git bisect with a test script, I got:
f798b1dda1c5de818b806189e523d1b75db7e72d is the first bad commit
That commit changes to using fs->gdsize rather than a calculated descriptor size.
The problem is that ext4fs_mount does not set fs->gdsize when revision_level is zero.
The simplest fix seemed to be to plug in the old calculation into the case when revision_level is zero, which made the problem go away. I know next to nothing about the internals ext2/3/4, so this may not be the right thing to do. I'm attaching a patch anyways:
The patch is also available at the following URL if the mailer has mangled it: https://ptpb.pw/KcQI
diff --git a/fs/ext4/ext4_common.c b/fs/ext4/ext4_common.c index bfebe7e..fa83586 100644 --- a/fs/ext4/ext4_common.c +++ b/fs/ext4/ext4_common.c @@ -2334,6 +2334,7 @@ int ext4fs_mount(unsigned part_length)
if (le32_to_cpu(data->sblock.revision_level) == 0) { fs->inodesz = 128; + fs->gdsize = EXT2_BLOCK_SIZE(data) / sizeof(struct ext2_block_group); } else { debug("EXT4 features COMPAT: %08x INCOMPAT: %08x RO_COMPAT: %08x\n", __le32_to_cpu(data->sblock.feature_compatibility),
Sent with [ProtonMail](https://protonmail.com) Secure Email.
On Sun, Dec 11, 2016 at 07:25:21PM -0500, FrostyBytes wrote:
I found that the latest U-Boot cannot handle filesystems generated by genext2fs. Attempts to list such a filesystem results in divide by zero.
Using git bisect with a test script, I got:
f798b1dda1c5de818b806189e523d1b75db7e72d is the first bad commit
That commit changes to using fs->gdsize rather than a calculated descriptor size.
The problem is that ext4fs_mount does not set fs->gdsize when revision_level is zero.
The simplest fix seemed to be to plug in the old calculation into the case when revision_level is zero, which made the problem go away. I know next to nothing about the internals ext2/3/4, so this may not be the right thing to do. I'm attaching a patch anyways:
The patch is also available at the following URL if the mailer has mangled it: https://ptpb.pw/KcQI
diff --git a/fs/ext4/ext4_common.c b/fs/ext4/ext4_common.c index bfebe7e..fa83586 100644 --- a/fs/ext4/ext4_common.c +++ b/fs/ext4/ext4_common.c @@ -2334,6 +2334,7 @@ int ext4fs_mount(unsigned part_length)
if (le32_to_cpu(data->sblock.revision_level) == 0) { fs->inodesz = 128;
- fs->gdsize = EXT2_BLOCK_SIZE(data) / sizeof(struct ext2_block_group);
} else { debug("EXT4 features COMPAT: %08x INCOMPAT: %08x RO_COMPAT: %08x\n", __le32_to_cpu(data->sblock.feature_compatibility),
Thanks for looking into this and providing the fix as well. Can you please reply back with a Signed-off-by and your real name? Thanks!
On Sonntag, 11. Dezember 2016 19:25:21 CET FrostyBytes wrote:
I found that the latest U-Boot cannot handle filesystems generated by genext2fs. Attempts to list such a filesystem results in divide by zero.
Using git bisect with a test script, I got:
f798b1dda1c5de818b806189e523d1b75db7e72d is the first bad commit
That commit changes to using fs->gdsize rather than a calculated descriptor size.
The problem is that ext4fs_mount does not set fs->gdsize when revision_level is zero.
The simplest fix seemed to be to plug in the old calculation into the case when revision_level is zero, which made the problem go away. I know next to nothing about the internals ext2/3/4, so this may not be the right thing to do. I'm attaching a patch anyways:
The patch is also available at the following URL if the mailer has mangled it: https://ptpb.pw/KcQI
diff --git a/fs/ext4/ext4_common.c b/fs/ext4/ext4_common.c index bfebe7e..fa83586 100644 --- a/fs/ext4/ext4_common.c +++ b/fs/ext4/ext4_common.c @@ -2334,6 +2334,7 @@ int ext4fs_mount(unsigned part_length)
if (le32_to_cpu(data->sblock.revision_level) == 0) { fs->inodesz = 128;
- fs->gdsize = EXT2_BLOCK_SIZE(data) / sizeof(struct ext2_block_group);
} else { debug("EXT4 features COMPAT: %08x INCOMPAT: %08x RO_COMPAT: %08x\n", __le32_to_cpu(data->sblock.feature_compatibility),
Although you are correct in that the initialization is missing here, your patch is wrong, and may even corrupt the filesystem.
Just make it fs->gdsize = 32;
Kind regards,
Stefan
participants (3)
-
FrostyBytes -
stefan.bruens＠rwth-aachen.de -
Tom Rini