On Tue, Jun 30, 2009 at 06:42:11PM +0200, Robert Millan wrote: > On Tue, Jun 30, 2009 at 04:42:50PM +0100, Colin Watson wrote: > > The simple fix is to call the BLKFLSBUF ioctl on any cached file > > descriptor open on the disk device after writing to the partition > > device. The attached patch (to be installed in > > debian/patches/cache_coherency.diff, with the obvious additional entry > > at the end of debian/patches/00list) does this. > > Hi Colin, > > You forgot to attach it.
D'oh. Really attached now. -- Colin Watson [[email protected]]
Description: Flush buffer cache on disk devices after writing to any partition devices they contain, otherwise we may read old data back. Debian: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=362028 Ubuntu: https://bugs.launchpad.net/bugs/185878 Index: b/stage2/disk_io.c =================================================================== --- a/stage2/disk_io.c +++ b/stage2/disk_io.c @@ -31,6 +31,11 @@ # include <device.h> #endif +#if defined(GRUB_UTIL) && defined(__linux__) +# include <sys/ioctl.h> /* ioctl */ +# include <linux/fs.h> /* BLKFLSBUF */ +#endif + /* instrumentation variables */ void (*disk_read_hook) (int, int, int) = NULL; void (*disk_read_func) (int, int, int) = NULL; @@ -366,6 +371,8 @@ devwrite (unsigned int sector, int sector_count, char *buf) { #if defined(GRUB_UTIL) && defined(__linux__) + int ret; + if (current_partition != 0xFFFFFF && is_disk_device (device_map, current_drive)) { @@ -373,8 +380,11 @@ embed a Stage 1.5 into a partition instead of a MBR, use system calls directly instead of biosdisk, because of the bug in Linux. *sigh* */ - return write_to_partition (device_map, current_drive, current_partition, - sector, sector_count, buf); + ret = write_to_partition (device_map, current_drive, current_partition, + sector, sector_count, buf); + if (ret && disks[current_drive].flags != -1) + ioctl (disks[current_drive].flags, BLKFLSBUF, 0); + return ret; } else #endif /* GRUB_UTIL && __linux__ */
