From cab229b2a679287710e4a23e4901ee020d0bf96a Mon Sep 17 00:00:00 2001 From: Scott Long Date: Mon, 20 Nov 2017 20:53:03 +0000 Subject: [PATCH] Update a comment in brelse() to match reality. --- sys/kern/vfs_bio.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c index 35d67d1b37b..4c223a828c1 100644 --- a/sys/kern/vfs_bio.c +++ b/sys/kern/vfs_bio.c @@ -2340,9 +2340,18 @@ brelse(struct buf *bp) !(bp->b_flags & B_INVAL)) { /* * Failed write, redirty. All errors except ENXIO (which - * means the device is gone) are expected to be potentially - * transient - underlying media might work if tried again - * after EIO, and memory might be available after an ENOMEM. + * means the device is gone) are treated as being + * transient. + * + * XXX Treating EIO as transient is not correct; the + * contract with the local storage device drivers is that + * they will only return EIO once the I/O is no longer + * retriable. Network I/O also respects this through the + * guarantees of TCP and/or the internal retries of NFS. + * ENOMEM might be transient, but we also have no way of + * knowing when its ok to retry/reschedule. In general, + * this entire case should be made obsolete through better + * error handling/recovery and resource scheduling. * * Do this also for buffers that failed with ENXIO, but have * non-empty dependencies - the soft updates code might need