Raphael S Carvalho
2013-Oct-15 20:40 UTC
[syslinux] Using *Syslinux 6.02* on *BTRFS* volumes corrupts the superblock.
BTRFS file systems will be corrupted on 6.02 (mainstream) and some previous versions. DYNAMIC_DEBUG makes the core larger than the allowed. Even before its introduction, building Syslinux with DEBUG_PORT enabled would probably corrupt BTRFS images (IIRC) *if* ldlinux.sys surpasses the maximum allowed size. Besides, the installer isn't safe at all... The check could be easily made there to avoid overwriting BTRFS superblocks. It simply install ldlinux.sys blindly even when it could be done otherwise (Not good in my opinion). I compiled Syslinux with several options in order to analyze how the size of ldlinux.sys behaves. Follow the log: DYNAMIC_DEBUG DISABLED & DEBUG_PORT DISABLED: 60474 bytes DYNAMIC_DEBUG ENABLED & DEBUG_PORT DISABLED: 67114 bytes DYNAMIC_DEBUG DISABLED & DEBUG_PORT ENABLED: 65673 bytes DYNAMIC_DEBUG ENABLED & DEBUG_PORT ENABLED: 65774 bytes PS: ldlinux.sys can currently be at most 64000 bytes (ADV_SIZE: 512). PS: Strangely, '*DYNAMIC_DEBUG ENABLED* WITH *DEBUG_PORT DISABLED*' makes the core larger than with both options enabled. Any comments on the matter *or not =) * would be really appreciated. -- Raphael S. Carvalho
H. Peter Anvin
2013-Oct-16 16:00 UTC
[syslinux] Using *Syslinux 6.02* on *BTRFS* volumes corrupts the superblock.
On 10/15/2013 01:40 PM, Raphael S Carvalho wrote:> BTRFS file systems will be corrupted on 6.02 (mainstream) and some > previous versions. > > DYNAMIC_DEBUG makes the core larger than the allowed. > Even before its introduction, building Syslinux with DEBUG_PORT > enabled would probably corrupt BTRFS images (IIRC) *if* ldlinux.sys > surpasses the maximum allowed size. > > Besides, the installer isn't safe at all... The check could be easily > made there to avoid overwriting BTRFS superblocks. It simply install > ldlinux.sys blindly even when it could be done otherwise (Not good in > my opinion). > > I compiled Syslinux with several options in order to analyze how the > size of ldlinux.sys behaves. Follow the log: > > DYNAMIC_DEBUG DISABLED & DEBUG_PORT DISABLED: 60474 bytes > DYNAMIC_DEBUG ENABLED & DEBUG_PORT DISABLED: 67114 bytes > DYNAMIC_DEBUG DISABLED & DEBUG_PORT ENABLED: 65673 bytes > DYNAMIC_DEBUG ENABLED & DEBUG_PORT ENABLED: 65774 bytes > > PS: ldlinux.sys can currently be at most 64000 bytes (ADV_SIZE: 512). > PS: Strangely, '*DYNAMIC_DEBUG ENABLED* WITH *DEBUG_PORT DISABLED*' > makes the core larger than with both options enabled. > > Any comments on the matter *or not =) * would be really appreciated. > -- > Raphael S. Carvalho >We should error out the build if ldlinux.sys is too large, this is something that is in the works I believe. But yes, an extra check in the installer would not be a bad thing. With DEBUG_PORT disabled, the debug output goes to the primary console, which requires a bit more code. -hpa
Seemingly Similar Threads
- [PATCH v2] core: Check size of ldlinux.sys at building time.
- [PATCH v3 1/7] lib: string: add functions to case-convert strings
- [PATCH v3 1/7] lib: string: add functions to case-convert strings
- "EDD Load error" on btrfs, how to debug?
- EFI: HP + syslinux = crash