From 2f21596882f4a0edc387051910d56f8732970080 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 15 Feb 2016 12:36:04 -0500 Subject: pNFS: Fix pnfs_mark_matching_lsegs_return() We don't need to schedule a layoutreturn if the layout segment can be freed immediately. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 482b6e94bb37..8d9fa4e40e89 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1740,6 +1740,16 @@ pnfs_set_plh_return_iomode(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode) lo->plh_return_iomode = iomode; } +/** + * pnfs_mark_matching_lsegs_return - Free or return matching layout segments + * @lo: pointer to layout header + * @tmp_list: list header to be used with pnfs_free_lseg_list() + * @return_range: describe layout segment ranges to be returned + * + * This function is mainly intended for use by layoutrecall. It attempts + * to free the layout segment immediately, or else to mark it for return + * as soon as its reference count drops to zero. + */ int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, @@ -1762,10 +1772,11 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, lseg, lseg->pls_range.iomode, lseg->pls_range.offset, lseg->pls_range.length); + if (mark_lseg_invalid(lseg, tmp_list)) + continue; + remaining++; set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags); pnfs_set_plh_return_iomode(lo, return_range->iomode); - if (!mark_lseg_invalid(lseg, tmp_list)) - remaining++; set_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags); } -- cgit v1.2.3 From e0fa0d0189f6be61343dbff1f71bef20125576e7 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 15 Feb 2016 12:56:17 -0500 Subject: pNFS: Always set NFS_LAYOUT_RETURN_REQUESTED with lo->plh_return_iomode When setting the layout return mode, we must always also set the NFS_LAYOUT_RETURN_REQUESTED flag to ensure that we send a layoutreturn. Otherwise pnfs_error_mark_layout_for_return() could set the mode, but fail to send the layoutreturn because another is already in flight. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 8d9fa4e40e89..5064f23ca8cc 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1738,6 +1738,7 @@ pnfs_set_plh_return_iomode(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode) if (lo->plh_return_iomode != 0) iomode = IOMODE_ANY; lo->plh_return_iomode = iomode; + set_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags); } /** @@ -1777,8 +1778,6 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, remaining++; set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags); pnfs_set_plh_return_iomode(lo, return_range->iomode); - set_bit(NFS_LAYOUT_RETURN_REQUESTED, - &lo->plh_flags); } return remaining; } -- cgit v1.2.3 From 2454dfea0aef8a883d4302c38124edb373d47a25 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 22 Feb 2016 17:34:59 -0500 Subject: NFSv4.x/pnfs: Fix a race between layoutget and pnfs_destroy_layout If the server reboots while there is a layoutget outstanding, then the call to pnfs_choose_layoutget_stateid() will fail with an EAGAIN error, which causes an infinite loop in send_layoutget(). The reason why we never break out of the loop is that the layout 'plh_block_lgets' field is never cleared. Fix is to replace plh_block_lgets with NFS_LAYOUT_INVALID_STID, which can be reset after a new layoutget. Fixes: ab7d763e477c5 ("pNFS: Ensure nfs4_layoutget_prepare returns...") Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 24 ++++++++++++++++++++++-- 1 file changed, 22 insertions(+), 2 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 5064f23ca8cc..8a41538f1155 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -252,6 +252,27 @@ pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo) } } +/* + * Mark a pnfs_layout_hdr and all associated layout segments as invalid + * + * In order to continue using the pnfs_layout_hdr, a full recovery + * is required. + * Note that caller must hold inode->i_lock. + */ +static void +pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo, + struct list_head *lseg_list) +{ + struct pnfs_layout_range range = { + .iomode = IOMODE_ANY, + .offset = 0, + .length = NFS4_MAX_UINT64, + }; + + set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags); + pnfs_mark_matching_lsegs_invalid(lo, lseg_list, &range); +} + static int pnfs_iomode_to_fail_bit(u32 iomode) { @@ -554,9 +575,8 @@ pnfs_destroy_layout(struct nfs_inode *nfsi) spin_lock(&nfsi->vfs_inode.i_lock); lo = nfsi->layout; if (lo) { - lo->plh_block_lgets++; /* permanently block new LAYOUTGETs */ - pnfs_mark_matching_lsegs_invalid(lo, &tmp_list, NULL); pnfs_get_layout_hdr(lo); + pnfs_mark_layout_stateid_invalid(lo, &tmp_list); pnfs_layout_clear_fail_bit(lo, NFS_LAYOUT_RO_FAILED); pnfs_layout_clear_fail_bit(lo, NFS_LAYOUT_RW_FAILED); spin_unlock(&nfsi->vfs_inode.i_lock); -- cgit v1.2.3 From 9fd4b9fc7695a4da1363a0b439853f88fb6a034f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 22 Feb 2016 17:46:34 -0500 Subject: NFSv4.x/pnfs: Fix a race between layoutget and bulk recalls Replace another case where the layout 'plh_block_lgets' can trigger infinite loops in send_layoutget(). Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 17 ++++++----------- 1 file changed, 6 insertions(+), 11 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 8a41538f1155..2fa483e6dbe2 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -259,7 +259,7 @@ pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo) * is required. * Note that caller must hold inode->i_lock. */ -static void +static int pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo, struct list_head *lseg_list) { @@ -270,7 +270,7 @@ pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo, }; set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags); - pnfs_mark_matching_lsegs_invalid(lo, lseg_list, &range); + return pnfs_mark_matching_lsegs_invalid(lo, lseg_list, &range); } static int @@ -637,11 +637,6 @@ pnfs_layout_free_bulk_destroy_list(struct list_head *layout_list, { struct pnfs_layout_hdr *lo; struct inode *inode; - struct pnfs_layout_range range = { - .iomode = IOMODE_ANY, - .offset = 0, - .length = NFS4_MAX_UINT64, - }; LIST_HEAD(lseg_list); int ret = 0; @@ -656,11 +651,11 @@ pnfs_layout_free_bulk_destroy_list(struct list_head *layout_list, spin_lock(&inode->i_lock); list_del_init(&lo->plh_bulk_destroy); - lo->plh_block_lgets++; /* permanently block new LAYOUTGETs */ - if (is_bulk_recall) - set_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); - if (pnfs_mark_matching_lsegs_invalid(lo, &lseg_list, &range)) + if (pnfs_mark_layout_stateid_invalid(lo, &lseg_list)) { + if (is_bulk_recall) + set_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); ret = -EAGAIN; + } spin_unlock(&inode->i_lock); pnfs_free_lseg_list(&lseg_list); /* Free all lsegs that are attached to commit buckets */ -- cgit v1.2.3