Pull ceph fixes from Ilya Dryomov: "A bunch of assorted fixes with the majority being hardening against malformed input and invalid data scenarios that don't happen in real deployments but can be utilized to trigger use-after-free and similar issues, some error path leak fixups and two patches from Max to avoid a potential hang in __ceph_get_caps() and unintended nesting of current->journal_info while handling replies from the MDS. All marked for stable" * tag 'ceph-for-7.2-rc5' of https://github.com/ceph/ceph-client: ceph: avoid fs reclaim while using current->journal_info ceph: add owner/capability checks for CEPH_IOC_SET_LAYOUT* ceph: fix hanging __ceph_get_caps() with stale mds_wanted rbd: Reset positive result codes to zero in object map update path libceph: bound pg_{temp,upmap,upmap_items} length to CEPH_PG_MAX_SIZE libceph: refresh auth->authorizer_buf{,_len} after authorizer update ceph: fix refcount leak in ceph_readdir() libceph: guard missing CRUSH type name lookup libceph: remove debugfs files before client teardown libceph: bound get_version reply decode to front len ceph: fix writeback_count leak in write_folio_nounlock() libceph: fix two unsafe bare decodes in decode_lockers() ceph: fix pre-auth out-of-bounds read on snaptrace in ceph_handle_caps() libceph: Reject monmaps advertising zero monitors libceph: reject zero bucket types in crush_decode libceph: Fix multiplication overflow in decode_new_up_state_weight()
Unified · Split
| @@ -1950,9 +1950,14 @@ static int rbd_object_map_update_finish(struct rbd_obj_request *obj_req, · 1949 unmodified lines | ||
| 1950 | 1950 | bool has_current_state; |
| 1951 | 1951 | void *p; |
| 1952 | 1952 | |
| 1953 | − if (osd_req->r_result) | |
| 1953 | + if (osd_req->r_result < 0) | |
| 1954 | 1954 | return osd_req->r_result; |
| 1955 | 1955 | |
| 1956 | + /* | |
| 1957 | + * Writes aren't allowed to return a data payload. | |
| 1958 | + */ | |
| 1959 | + WARN_ON_ONCE(osd_req->r_result > 0); | |
| 1960 | + | |
| 1956 | 1961 | /* |
| 1957 | 1962 | * Nothing to do for a snapshot object map. |
| 1958 | 1963 | */ |
| @@ -790,6 +790,9 @@ static int write_folio_nounlock(struct folio *folio, · 789 unmodified lines | ||
| 790 | 790 | ceph_wbc.truncate_size, true); |
| 791 | 791 | if (IS_ERR(req)) { |
| 792 | 792 | folio_redirty_for_writepage(wbc, folio); |
| 793 | + if (atomic_long_dec_return(&fsc->writeback_count) < | |
| 794 | + CONGESTION_OFF_THRESH(fsc->mount_options->congestion_kb)) | |
| 795 | + fsc->write_congested = false; | |
| 793 | 796 | return PTR_ERR(req); |
| 794 | 797 | } |
| 795 | 798 | |
| @@ -809,6 +812,9 @@ static int write_folio_nounlock(struct folio *folio, · 13 unmodified lines | ||
| 809 | 812 | folio_redirty_for_writepage(wbc, folio); |
| 810 | 813 | folio_end_writeback(folio); |
| 811 | 814 | ceph_osdc_put_request(req); |
| 815 | + if (atomic_long_dec_return(&fsc->writeback_count) < | |
| 816 | + CONGESTION_OFF_THRESH(fsc->mount_options->congestion_kb)) | |
| 817 | + fsc->write_congested = false; | |
| 812 | 818 | return PTR_ERR(bounce_page); |
| 813 | 819 | } |
| 814 | 820 | } |
| @@ -847,6 +853,9 @@ static int write_folio_nounlock(struct folio *folio, · 32 unmodified lines | ||
| 847 | 853 | ceph_vinop(inode), folio); |
| 848 | 854 | folio_redirty_for_writepage(wbc, folio); |
| 849 | 855 | folio_end_writeback(folio); |
| 856 | + if (atomic_long_dec_return(&fsc->writeback_count) < | |
| 857 | + CONGESTION_OFF_THRESH(fsc->mount_options->congestion_kb)) | |
| 858 | + fsc->write_congested = false; | |
| 850 | 859 | return err; |
| 851 | 860 | } |
| 852 | 861 | if (err == -EBLOCKLISTED) |
| @@ -3094,7 +3094,19 @@ int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi, int need, · 3093 unmodified lines | ||
| 3094 | 3094 | ret = -ERESTARTSYS; |
| 3095 | 3095 | break; |
| 3096 | 3096 | } |
| 3097 | − wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); | |
| 3097 | + | |
| 3098 | + /* | |
| 3099 | + * If a cap update is lost after | |
| 3100 | + * mds_wanted was raised, waiting | |
| 3101 | + * forever will never make progress. | |
| 3102 | + * Retry the renew path periodically | |
| 3103 | + * so we can resend synchronously. | |
| 3104 | + */ | |
| 3105 | + if (!wait_woken(&wait, TASK_INTERRUPTIBLE, | |
| 3106 | + CEPH_GET_CAPS_WAIT_TIMEOUT)) { | |
| 3107 | + ret = -EUCLEAN; | |
| 3108 | + break; | |
| 3109 | + } | |
| 3098 | 3110 | } |
| 3099 | 3111 | |
| 3100 | 3112 | remove_wait_queue(&ci->i_cap_wq, &wait); |
| @@ -3128,7 +3140,8 @@ int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi, int need, · 27 unmodified lines | ||
| 3128 | 3140 | continue; |
| 3129 | 3141 | } |
| 3130 | 3142 | if (ret == -EUCLEAN) { |
| 3131 | − /* session was killed, try renew caps */ | |
| 3143 | + /* session was killed or a waited cap | |
| 3144 | + * request needs a retry */ | |
| 3132 | 3145 | ret = ceph_renew_caps(inode, flags); |
| 3133 | 3146 | if (ret == 0) |
| 3134 | 3147 | continue; |
| @@ -4375,6 +4388,7 @@ void ceph_handle_caps(struct ceph_mds_session *session, · 1240 unmodified lines | ||
| 4375 | 4388 | |
| 4376 | 4389 | snaptrace = h + 1; |
| 4377 | 4390 | snaptrace_len = le32_to_cpu(h->snap_trace_len); |
| 4391 | + ceph_decode_need(&snaptrace, end, snaptrace_len, bad); | |
| 4378 | 4392 | p = snaptrace + snaptrace_len; |
| 4379 | 4393 | |
| 4380 | 4394 | if (msg_version >= 2) { |
| @@ -546,11 +546,16 @@ more: · 545 unmodified lines | ||
| 546 | 546 | pr_warn_client(cl, |
| 547 | 547 | "%p %llx.%llx rde->offset 0x%llx ctx->pos 0x%llx\n", |
| 548 | 548 | inode, ceph_vinop(inode), rde->offset, ctx->pos); |
| 549 | + ceph_mdsc_put_request(dfi->last_readdir); | |
| 550 | + dfi->last_readdir = NULL; | |
| 549 | 551 | return -EIO; |
| 550 | 552 | } |
| 551 | 553 | |
| 552 | − if (WARN_ON_ONCE(!rde->inode.in)) | |
| 554 | + if (WARN_ON_ONCE(!rde->inode.in)) { | |
| 555 | + ceph_mdsc_put_request(dfi->last_readdir); | |
| 556 | + dfi->last_readdir = NULL; | |
| 553 | 557 | return -EIO; |
| 558 | + } | |
| 554 | 559 | |
| 555 | 560 | ctx->pos = rde->offset; |
| 556 | 561 | doutc(cl, "%p %llx.%llx (%d/%d) -> %llx '%.*s' %p\n", inode, |
| @@ -314,7 +314,7 @@ static int ceph_init_file(struct inode *inode, struct file *file, int fmode) · 313 unmodified lines | ||
| 314 | 314 | } |
| 315 | 315 | |
| 316 | 316 | /* |
| 317 | − * try renew caps after session gets killed. | |
| 317 | + * Retry cap acquisition after a stale session or a lost cap update. | |
| 318 | 318 | */ |
| 319 | 319 | int ceph_renew_caps(struct inode *inode, int fmode) |
| 320 | 320 | { |
| @@ -322,14 +322,15 @@ int ceph_renew_caps(struct inode *inode, int fmode) · 1 unmodified line | ||
| 322 | 322 | struct ceph_client *cl = mdsc->fsc->client; |
| 323 | 323 | struct ceph_inode_info *ci = ceph_inode(inode); |
| 324 | 324 | struct ceph_mds_request *req; |
| 325 | − int err, flags, wanted; | |
| 325 | + int err, flags, wanted, issued; | |
| 326 | 326 | |
| 327 | 327 | spin_lock(&ci->i_ceph_lock); |
| 328 | 328 | __ceph_touch_fmode(ci, mdsc, fmode); |
| 329 | 329 | wanted = __ceph_caps_file_wanted(ci); |
| 330 | + issued = __ceph_caps_issued(ci, NULL); | |
| 330 | 331 | if (__ceph_is_any_real_caps(ci) && |
| 331 | − (!(wanted & CEPH_CAP_ANY_WR) || ci->i_auth_cap)) { | |
| 332 | − int issued = __ceph_caps_issued(ci, NULL); | |
| 332 | + (!(wanted & CEPH_CAP_ANY_WR) || ci->i_auth_cap) && | |
| 333 | + (issued & wanted) == wanted) { | |
| 333 | 334 | spin_unlock(&ci->i_ceph_lock); |
| 334 | 335 | doutc(cl, "%p %llx.%llx want %s issued %s updating mds_wanted\n", |
| 335 | 336 | inode, ceph_vinop(inode), ceph_cap_string(wanted), |
| @@ -72,6 +72,9 @@ static long ceph_ioctl_set_layout(struct file *file, void __user *arg) · 71 unmodified lines | ||
| 72 | 72 | struct ceph_ioctl_layout nl; |
| 73 | 73 | int err; |
| 74 | 74 | |
| 75 | + if (!inode_owner_or_capable(&nop_mnt_idmap, inode)) | |
| 76 | + return -EACCES; | |
| 77 | + | |
| 75 | 78 | if (copy_from_user(&l, arg, sizeof(l))) |
| 76 | 79 | return -EFAULT; |
| 77 | 80 | |
| @@ -142,6 +145,9 @@ static long ceph_ioctl_set_layout_policy (struct file *file, void __user *arg) · 64 unmodified lines | ||
| 142 | 145 | int err; |
| 143 | 146 | struct ceph_mds_client *mdsc = ceph_sb_to_fs_client(inode->i_sb)->mdsc; |
| 144 | 147 | |
| 148 | + if (!inode_owner_or_capable(&nop_mnt_idmap, inode)) | |
| 149 | + return -EACCES; | |
| 150 | + | |
| 145 | 151 | /* copy and validate */ |
| 146 | 152 | if (copy_from_user(&l, arg, sizeof(l))) |
| 147 | 153 | return -EFAULT; |
| @@ -6,6 +6,7 @@ · 5 unmodified lines | ||
| 6 | 6 | #include <linux/slab.h> |
| 7 | 7 | #include <linux/gfp.h> |
| 8 | 8 | #include <linux/sched.h> |
| 9 | +#include <linux/sched/mm.h> | |
| 9 | 10 | #include <linux/delay.h> |
| 10 | 11 | #include <linux/debugfs.h> |
| 11 | 12 | #include <linux/seq_file.h> |
| @@ -4015,6 +4016,7 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg) · 4003 unmodified lines | ||
| 4015 | 4016 | struct ceph_mds_reply_head *head = msg->front.iov_base; |
| 4016 | 4017 | struct ceph_mds_reply_info_parsed *rinfo; /* parsed reply info */ |
| 4017 | 4018 | struct ceph_snap_realm *realm; |
| 4019 | + unsigned int nofs_flags; | |
| 4018 | 4020 | u64 tid; |
| 4019 | 4021 | int err, result; |
| 4020 | 4022 | int mds = session->s_mds; |
| @@ -4158,6 +4160,14 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg) · 137 unmodified lines | ||
| 4158 | 4160 | |
| 4159 | 4161 | /* insert trace into our cache */ |
| 4160 | 4162 | mutex_lock(&req->r_fill_mutex); |
| 4163 | + | |
| 4164 | + /* disable fs reclaim while we are using current->journal_info | |
| 4165 | + * for our own purposes, or else shrinkers of other | |
| 4166 | + * filesystems might dereference this pointer as a different | |
| 4167 | + * type | |
| 4168 | + */ | |
| 4169 | + nofs_flags = memalloc_nofs_save(); | |
| 4170 | + | |
| 4161 | 4171 | current->journal_info = req; |
| 4162 | 4172 | err = ceph_fill_trace(mdsc->fsc->sb, req); |
| 4163 | 4173 | if (err == 0) { |
| @@ -4166,6 +4176,7 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg) · 2 unmodified lines | ||
| 4166 | 4176 | err = ceph_readdir_prepopulate(req, req->r_session); |
| 4167 | 4177 | } |
| 4168 | 4178 | current->journal_info = NULL; |
| 4179 | + memalloc_nofs_restore(nofs_flags); | |
| 4169 | 4180 | mutex_unlock(&req->r_fill_mutex); |
| 4170 | 4181 | |
| 4171 | 4182 | up_read(&mdsc->snap_rwsem); |
| @@ -77,6 +77,7 @@ struct ceph_fs_client; · 76 unmodified lines | ||
| 77 | 77 | struct ceph_cap; |
| 78 | 78 | |
| 79 | 79 | #define MDS_AUTH_UID_ANY -1 |
| 80 | +#define CEPH_GET_CAPS_WAIT_TIMEOUT (5 * HZ) | |
| 80 | 81 | #define CEPH_CAP_FLUSH_WAIT_TIMEOUT_SEC 60 |
| 81 | 82 | #define CEPH_CAP_FLUSH_MAX_DUMP_ENTRIES 5 |
| 82 | 83 | #define CEPH_CAP_FLUSH_MAX_DUMP_ITERS 5 |
| @@ -849,9 +849,16 @@ static int ceph_x_update_authorizer( · 848 unmodified lines | ||
| 849 | 849 | |
| 850 | 850 | au = (struct ceph_x_authorizer *)auth->authorizer; |
| 851 | 851 | if (au->secret_id < th->secret_id) { |
| 852 | + int ret; | |
| 853 | + | |
| 852 | 854 | dout("ceph_x_update_authorizer service %u secret %llu < %llu\n", |
| 853 | 855 | au->service, au->secret_id, th->secret_id); |
| 854 | − return ceph_x_build_authorizer(ac, th, au); | |
| 856 | + ret = ceph_x_build_authorizer(ac, th, au); | |
| 857 | + if (ret) | |
| 858 | + return ret; | |
| 859 | + | |
| 860 | + auth->authorizer_buf = au->buf->vec.iov_base; | |
| 861 | + auth->authorizer_buf_len = au->buf->vec.iov_len; | |
| 855 | 862 | } |
| 856 | 863 | return 0; |
| 857 | 864 | } |
| @@ -762,13 +762,13 @@ void ceph_destroy_client(struct ceph_client *client) · 761 unmodified lines | ||
| 762 | 762 | |
| 763 | 763 | atomic_set(&client->msgr.stopping, 1); |
| 764 | 764 | |
| 765 | + ceph_debugfs_client_cleanup(client); | |
| 766 | + | |
| 765 | 767 | /* unmount */ |
| 766 | 768 | ceph_osdc_stop(&client->osdc); |
| 767 | 769 | ceph_monc_stop(&client->monc); |
| 768 | 770 | ceph_messenger_fini(&client->msgr); |
| 769 | 771 | |
| 770 | − ceph_debugfs_client_cleanup(client); | |
| 771 | − | |
| 772 | 772 | ceph_destroy_options(client->options); |
| 773 | 773 | |
| 774 | 774 | kfree(client); |
| @@ -299,7 +299,7 @@ static int decode_lockers(void **p, void *end, u8 *type, char **tag, · 298 unmodified lines | ||
| 299 | 299 | if (ret) |
| 300 | 300 | return ret; |
| 301 | 301 | |
| 302 | − *num_lockers = ceph_decode_32(p); | |
| 302 | + ceph_decode_32_safe(p, end, *num_lockers, err_inval); | |
| 303 | 303 | *lockers = kzalloc_objs(**lockers, *num_lockers, GFP_NOIO); |
| 304 | 304 | if (!*lockers) |
| 305 | 305 | return -ENOMEM; |
| @@ -310,7 +310,8 @@ static int decode_lockers(void **p, void *end, u8 *type, char **tag, · 4 unmodified lines | ||
| 310 | 310 | goto err_free_lockers; |
| 311 | 311 | } |
| 312 | 312 | |
| 313 | − *type = ceph_decode_8(p); | |
| 313 | + ret = -EINVAL; | |
| 314 | + ceph_decode_8_safe(p, end, *type, err_free_lockers); | |
| 314 | 315 | s = ceph_extract_encoded_string(p, end, NULL, GFP_NOIO); |
| 315 | 316 | if (IS_ERR(s)) { |
| 316 | 317 | ret = PTR_ERR(s); |
| @@ -320,6 +321,9 @@ static int decode_lockers(void **p, void *end, u8 *type, char **tag, · 3 unmodified lines | ||
| 320 | 321 | *tag = s; |
| 321 | 322 | return 0; |
| 322 | 323 | |
| 324 | +err_inval: | |
| 325 | + return -EINVAL; | |
| 326 | + | |
| 323 | 327 | err_free_lockers: |
| 324 | 328 | ceph_free_lockers(*lockers, *num_lockers); |
| 325 | 329 | return ret; |
| @@ -114,7 +114,7 @@ static struct ceph_monmap *ceph_monmap_decode(void **p, void *end, bool msgr2) · 113 unmodified lines | ||
| 114 | 114 | |
| 115 | 115 | dout("%s fsid %pU epoch %u num_mon %u\n", __func__, &fsid, epoch, |
| 116 | 116 | num_mon); |
| 117 | − if (num_mon > CEPH_MAX_MON) | |
| 117 | + if (num_mon == 0 || num_mon > CEPH_MAX_MON) | |
| 118 | 118 | goto e_inval; |
| 119 | 119 | |
| 120 | 120 | monmap = kmalloc_flex(*monmap, mon_inst, num_mon, GFP_NOIO); |
| @@ -821,7 +821,7 @@ static void handle_get_version_reply(struct ceph_mon_client *monc, · 700 unmodified lines | ||
| 821 | 821 | struct ceph_mon_generic_request *req; |
| 822 | 822 | u64 tid = le64_to_cpu(msg->hdr.tid); |
| 823 | 823 | void *p = msg->front.iov_base; |
| 824 | − void *end = p + msg->front_alloc_len; | |
| 824 | + void *const end = p + msg->front.iov_len; | |
| 825 | 825 | u64 handle; |
| 826 | 826 | |
| 827 | 827 | dout("%s msg %p tid %llu\n", __func__, msg, tid); |
| @@ -518,6 +518,8 @@ static struct crush_map *crush_decode(void *pbyval, void *end) · 517 unmodified lines | ||
| 518 | 518 | ceph_decode_need(p, end, 4*sizeof(u32), bad); |
| 519 | 519 | b->id = ceph_decode_32(p); |
| 520 | 520 | b->type = ceph_decode_16(p); |
| 521 | + if (b->type == 0) | |
| 522 | + goto bad; | |
| 521 | 523 | b->alg = ceph_decode_8(p); |
| 522 | 524 | if (b->alg != alg) { |
| 523 | 525 | b->alg = 0; |
| @@ -1436,7 +1438,7 @@ static struct ceph_pg_mapping *__decode_pg_temp(void **p, void *end, · 912 unmodified lines | ||
| 1436 | 1438 | ceph_decode_32_safe(p, end, len, e_inval); |
| 1437 | 1439 | if (len == 0 && incremental) |
| 1438 | 1440 | return NULL; /* new_pg_temp: [] to remove */ |
| 1439 | − if ((size_t)len > (SIZE_MAX - sizeof(*pg)) / sizeof(u32)) | |
| 1441 | + if (len > CEPH_PG_MAX_SIZE) | |
| 1440 | 1442 | return ERR_PTR(-EINVAL); |
| 1441 | 1443 | |
| 1442 | 1444 | ceph_decode_need(p, end, len * sizeof(u32), e_inval); |
| @@ -1617,7 +1619,7 @@ static struct ceph_pg_mapping *__decode_pg_upmap_items(void **p, void *end, · 174 unmodified lines | ||
| 1617 | 1619 | u32 len, i; |
| 1618 | 1620 | |
| 1619 | 1621 | ceph_decode_32_safe(p, end, len, e_inval); |
| 1620 | − if ((size_t)len > (SIZE_MAX - sizeof(*pg)) / (2 * sizeof(u32))) | |
| 1622 | + if (len > CEPH_PG_MAX_SIZE) | |
| 1621 | 1623 | return ERR_PTR(-EINVAL); |
| 1622 | 1624 | |
| 1623 | 1625 | ceph_decode_need(p, end, 2 * len * sizeof(u32), e_inval); |
| @@ -1842,6 +1844,8 @@ static int decode_new_up_state_weight(void **p, void *end, u8 struct_v, · 218 unmodified lines | ||
| 1842 | 1844 | void *new_up_client; |
| 1843 | 1845 | void *new_state; |
| 1844 | 1846 | void *new_weight_end; |
| 1847 | + const u32 new_state_item_size = | |
| 1848 | + sizeof(u32) + (struct_v >= 5 ? sizeof(u32) : sizeof(u8)); | |
| 1845 | 1849 | u32 len; |
| 1846 | 1850 | int ret; |
| 1847 | 1851 | int i; |
| @@ -1862,7 +1866,8 @@ static int decode_new_up_state_weight(void **p, void *end, u8 struct_v, · 14 unmodified lines | ||
| 1862 | 1866 | |
| 1863 | 1867 | new_state = *p; |
| 1864 | 1868 | ceph_decode_32_safe(p, end, len, e_inval); |
| 1865 | − len *= sizeof(u32) + (struct_v >= 5 ? sizeof(u32) : sizeof(u8)); | |
| 1869 | + if (check_mul_overflow(len, new_state_item_size, &len)) | |
| 1870 | + goto e_inval; | |
| 1866 | 1871 | ceph_decode_need(p, end, len, e_inval); |
| 1867 | 1872 | *p += len; |
| 1868 | 1873 | |
| @@ -3055,8 +3060,11 @@ static int get_immediate_parent(struct crush_map *c, int id, · 1186 unmodified lines | ||
| 3055 | 3060 | if (b->items[j] != id) |
| 3056 | 3061 | continue; |
| 3057 | 3062 | |
| 3058 | − *parent_type_id = b->type; | |
| 3059 | 3063 | type_cn = lookup_crush_name(&c->type_names, b->type); |
| 3064 | + if (WARN_ON_ONCE(!type_cn)) | |
| 3065 | + continue; | |
| 3066 | + | |
| 3067 | + *parent_type_id = b->type; | |
| 3060 | 3068 | parent_loc->cl_type_name = type_cn->cn_name; |
| 3061 | 3069 | parent_loc->cl_name = cn->cn_name; |
| 3062 | 3070 | return b->id; |