From a1f26210f355eb26f59e445ca4043c9413231cbb Mon Sep 17 00:00:00 2001 From: Warner Losh Date: Sat, 31 Oct 2015 04:53:07 +0000 Subject: [PATCH] The error classification from lower layers is a poor indicator of whether an error is recoverable. Always re-dirty the buffer on errors from write requests. The invalidation we used to do for errors not EIO doesn't need to be done for a device that's really gone, since that's done in a different path. Reviewed by: mckusick@, kib@ --- sys/kern/vfs_bio.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c index 0f1f882eeab..1e51db39543 100644 --- a/sys/kern/vfs_bio.c +++ b/sys/kern/vfs_bio.c @@ -2266,19 +2266,17 @@ brelse(struct buf *bp) bdirty(bp); } if (bp->b_iocmd == BIO_WRITE && (bp->b_ioflags & BIO_ERROR) && - bp->b_error == EIO && !(bp->b_flags & B_INVAL)) { + !(bp->b_flags & B_INVAL)) { /* * Failed write, redirty. Must clear BIO_ERROR to prevent - * pages from being scrapped. If the error is anything - * other than an I/O error (EIO), assume that retrying - * is futile. + * pages from being scrapped. */ bp->b_ioflags &= ~BIO_ERROR; bdirty(bp); } else if ((bp->b_flags & (B_NOCACHE | B_INVAL)) || (bp->b_ioflags & BIO_ERROR) || (bp->b_bufsize <= 0)) { /* - * Either a failed I/O or we were asked to free or not + * Either a failed read I/O or we were asked to free or not * cache the buffer. */ bp->b_flags |= B_INVAL;