ну тогда и я ещё раз :-)  оттуда же:
ext3 recovery:
For most detected errors, ext3 propagates the error to the user (RPropagate). For
read failures, ext3 also often aborts the journal (RStop); aborting the journal usually
leads to a read-only remount of the file system, preventing future updates without
explicit administrator interaction. Ext3 also uses retry (RRetry), although sparingly;
when a prefetch read fails, ext3 retries only the originally requested block.
ReiserFS recovery:
The most prominent aspect of the recovery policy of ReiserFS is its tendency to
panic the system upon detection of virtually any write failure (RStop). When
ReiserFS calls panic, the file system crashes, usually leading to a reboot and
recovery sequence. By doing so, ReiserFS attempts to ensure that its on-disk structures
are not corrupted. ReiserFS recovers from read and write failures differently.
For most read failures, ReiserFS propagates the error to the user (RPropagate) and
sometimes performs a single retry (RRetry) (e.g., when a data block read fails, or
when an indirect block read fails during unlink, truncate, and write operations).
However, ReiserFS never retries upon a write failure.