From 02b2f549d502b46e68b97ea1452fb8853b3327dd Mon Sep 17 00:00:00 2001 From: Dongsheng Yang Date: Tue, 18 Dec 2018 04:31:48 -0500 Subject: [PATCH 1/3] libceph: allow setting abort_on_full for rbd Introduce a new option abort_on_full, default to false. Then we can get -ENOSPC when the pool is full, or reaches quota. [ Don't show abort_on_full in /proc/mounts. ] Signed-off-by: Dongsheng Yang Reviewed-by: Ilya Dryomov Signed-off-by: Ilya Dryomov --- fs/ceph/super.c | 4 ++-- include/linux/ceph/libceph.h | 6 ++++-- include/linux/ceph/osd_client.h | 1 - net/ceph/ceph_common.c | 11 ++++++++++- net/ceph/debugfs.c | 2 +- net/ceph/osd_client.c | 4 ++-- 6 files changed, 19 insertions(+), 9 deletions(-) diff --git a/fs/ceph/super.c b/fs/ceph/super.c index 4e9a7cc488da..da2cd8e89062 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -530,7 +530,7 @@ static int ceph_show_options(struct seq_file *m, struct dentry *root) seq_putc(m, ','); pos = m->count; - ret = ceph_print_client_options(m, fsc->client); + ret = ceph_print_client_options(m, fsc->client, false); if (ret) return ret; @@ -640,7 +640,7 @@ static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt, opt = NULL; /* fsc->client now owns this */ fsc->client->extra_mon_dispatch = extra_mon_dispatch; - fsc->client->osdc.abort_on_full = true; + ceph_set_opt(fsc->client, ABORT_ON_FULL); if (!fsopt->mds_namespace) { ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP, diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h index 68bb09c29ce8..a420c07904bc 100644 --- a/include/linux/ceph/libceph.h +++ b/include/linux/ceph/libceph.h @@ -35,6 +35,7 @@ #define CEPH_OPT_NOMSGAUTH (1<<4) /* don't require msg signing feat */ #define CEPH_OPT_TCP_NODELAY (1<<5) /* TCP_NODELAY on TCP sockets */ #define CEPH_OPT_NOMSGSIGN (1<<6) /* don't sign msgs */ +#define CEPH_OPT_ABORT_ON_FULL (1<<7) /* abort w/ ENOSPC when full */ #define CEPH_OPT_DEFAULT (CEPH_OPT_TCP_NODELAY) @@ -53,7 +54,7 @@ struct ceph_options { unsigned long osd_request_timeout; /* jiffies */ /* - * any type that can't be simply compared or doesn't need need + * any type that can't be simply compared or doesn't need * to be compared should go beyond this point, * ceph_compare_options() should be updated accordingly */ @@ -281,7 +282,8 @@ extern struct ceph_options *ceph_parse_options(char *options, const char *dev_name, const char *dev_name_end, int (*parse_extra_token)(char *c, void *private), void *private); -int ceph_print_client_options(struct seq_file *m, struct ceph_client *client); +int ceph_print_client_options(struct seq_file *m, struct ceph_client *client, + bool show_all); extern void ceph_destroy_options(struct ceph_options *opt); extern int ceph_compare_options(struct ceph_options *new_opt, struct ceph_client *client); diff --git a/include/linux/ceph/osd_client.h b/include/linux/ceph/osd_client.h index 7a2af5034278..2294f963dab7 100644 --- a/include/linux/ceph/osd_client.h +++ b/include/linux/ceph/osd_client.h @@ -354,7 +354,6 @@ struct ceph_osd_client { struct rb_root linger_map_checks; atomic_t num_requests; atomic_t num_homeless; - bool abort_on_full; /* abort w/ ENOSPC when full */ int abort_err; struct delayed_work timeout_work; struct delayed_work osds_timeout_work; diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c index 87afb9ec4c68..9cab80207ced 100644 --- a/net/ceph/ceph_common.c +++ b/net/ceph/ceph_common.c @@ -255,6 +255,7 @@ enum { Opt_nocephx_sign_messages, Opt_tcp_nodelay, Opt_notcp_nodelay, + Opt_abort_on_full, }; static match_table_t opt_tokens = { @@ -280,6 +281,7 @@ static match_table_t opt_tokens = { {Opt_nocephx_sign_messages, "nocephx_sign_messages"}, {Opt_tcp_nodelay, "tcp_nodelay"}, {Opt_notcp_nodelay, "notcp_nodelay"}, + {Opt_abort_on_full, "abort_on_full"}, {-1, NULL} }; @@ -535,6 +537,10 @@ ceph_parse_options(char *options, const char *dev_name, opt->flags &= ~CEPH_OPT_TCP_NODELAY; break; + case Opt_abort_on_full: + opt->flags |= CEPH_OPT_ABORT_ON_FULL; + break; + default: BUG_ON(token); } @@ -549,7 +555,8 @@ out: } EXPORT_SYMBOL(ceph_parse_options); -int ceph_print_client_options(struct seq_file *m, struct ceph_client *client) +int ceph_print_client_options(struct seq_file *m, struct ceph_client *client, + bool show_all) { struct ceph_options *opt = client->options; size_t pos = m->count; @@ -574,6 +581,8 @@ int ceph_print_client_options(struct seq_file *m, struct ceph_client *client) seq_puts(m, "nocephx_sign_messages,"); if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0) seq_puts(m, "notcp_nodelay,"); + if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL)) + seq_puts(m, "abort_on_full,"); if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT) seq_printf(m, "mount_timeout=%d,", diff --git a/net/ceph/debugfs.c b/net/ceph/debugfs.c index 02952605d121..46f65709a6ff 100644 --- a/net/ceph/debugfs.c +++ b/net/ceph/debugfs.c @@ -375,7 +375,7 @@ static int client_options_show(struct seq_file *s, void *p) struct ceph_client *client = s->private; int ret; - ret = ceph_print_client_options(s, client); + ret = ceph_print_client_options(s, client, true); if (ret) return ret; diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index d23a9f81f3d7..fa9530dd876e 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -2315,7 +2315,7 @@ again: (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || pool_full(osdc, req->r_t.base_oloc.pool))) { dout("req %p full/pool_full\n", req); - if (osdc->abort_on_full) { + if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) { err = -ENOSPC; } else { pr_warn_ratelimited("FULL or reached pool quota\n"); @@ -2545,7 +2545,7 @@ static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc) { bool victims = false; - if (osdc->abort_on_full && + if (ceph_test_opt(osdc->client, ABORT_ON_FULL) && (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc))) for_each_request(osdc, abort_on_full_fn, &victims); } From c64a2b0516a02361d8deb1f038647c29020d0852 Mon Sep 17 00:00:00 2001 From: Souptick Joarder Date: Sat, 5 Jan 2019 01:00:29 +0530 Subject: [PATCH 2/3] ceph: use vmf_error() in ceph_filemap_fault() This code is converted to use vmf_error(). Signed-off-by: Souptick Joarder Reviewed-by: Ilya Dryomov Signed-off-by: Ilya Dryomov --- fs/ceph/addr.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 5d0c05e288cc..a47c541f8006 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -1494,10 +1494,7 @@ static vm_fault_t ceph_filemap_fault(struct vm_fault *vmf) if (err < 0 || off >= i_size_read(inode)) { unlock_page(page); put_page(page); - if (err == -ENOMEM) - ret = VM_FAULT_OOM; - else - ret = VM_FAULT_SIGBUS; + ret = vmf_error(err); goto out_inline; } if (err < PAGE_SIZE) From 85f5a4d666fd9be73856ed16bb36c5af5b406b29 Mon Sep 17 00:00:00 2001 From: Ilya Dryomov Date: Tue, 8 Jan 2019 19:47:38 +0100 Subject: [PATCH 3/3] rbd: don't return 0 on unmap if RBD_DEV_FLAG_REMOVING is set There is a window between when RBD_DEV_FLAG_REMOVING is set and when the device is removed from rbd_dev_list. During this window, we set "already" and return 0. Returning 0 from write(2) can confuse userspace tools because 0 indicates that nothing was written. In particular, "rbd unmap" will retry the write multiple times a second: 10:28:05.463299 write(4, "0", 1) = 0 10:28:05.463509 write(4, "0", 1) = 0 10:28:05.463720 write(4, "0", 1) = 0 10:28:05.463942 write(4, "0", 1) = 0 10:28:05.464155 write(4, "0", 1) = 0 Cc: stable@vger.kernel.org Signed-off-by: Ilya Dryomov Tested-by: Dongsheng Yang --- drivers/block/rbd.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 8e5140bbf241..1e92b61d0bd5 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -5986,7 +5986,6 @@ static ssize_t do_rbd_remove(struct bus_type *bus, struct list_head *tmp; int dev_id; char opt_buf[6]; - bool already = false; bool force = false; int ret; @@ -6019,13 +6018,13 @@ static ssize_t do_rbd_remove(struct bus_type *bus, spin_lock_irq(&rbd_dev->lock); if (rbd_dev->open_count && !force) ret = -EBUSY; - else - already = test_and_set_bit(RBD_DEV_FLAG_REMOVING, - &rbd_dev->flags); + else if (test_and_set_bit(RBD_DEV_FLAG_REMOVING, + &rbd_dev->flags)) + ret = -EINPROGRESS; spin_unlock_irq(&rbd_dev->lock); } spin_unlock(&rbd_dev_list_lock); - if (ret < 0 || already) + if (ret) return ret; if (force) {