diff options
Diffstat (limited to 'fs')
33 files changed, 1308 insertions, 370 deletions
diff --git a/fs/ntfs/attrib.c b/fs/ntfs/attrib.c index 848a0d338b89..c949ff765075 100644 --- a/fs/ntfs/attrib.c +++ b/fs/ntfs/attrib.c @@ -2917,7 +2917,7 @@ retry: attr_ni = NULL; /* Allocate new extent. */ - err = ntfs_mft_record_alloc(ni->vol, 0, &attr_ni, ni, NULL); + err = ntfs_mft_record_alloc(ni->vol, 0, &attr_ni, ni, NULL, -1); if (err) { ntfs_error(sb, "Failed to allocate extent record"); goto err_out; @@ -3550,7 +3550,7 @@ int ntfs_attr_record_move_away(struct ntfs_attr_search_ctx *ctx, int extra) * new extent and move attribute to it. */ ni = NULL; - err = ntfs_mft_record_alloc(base_ni->vol, 0, &ni, base_ni, NULL); + err = ntfs_mft_record_alloc(base_ni->vol, 0, &ni, base_ni, NULL, -1); if (err) { ntfs_error(sb, "Couldn't allocate MFT record, err : %d", err); return err; @@ -3574,7 +3574,8 @@ int ntfs_attr_record_move_away(struct ntfs_attr_search_ctx *ctx, int extra) * update allocated and compressed size. */ static int ntfs_attr_update_meta(struct attr_record *a, struct ntfs_inode *ni, - struct mft_record *m, struct ntfs_attr_search_ctx *ctx) + struct mft_record *m, struct ntfs_attr_search_ctx *ctx, + struct ntfs_inode *locked_ni, bool defer_attrlist) { int sparse, err = 0; struct ntfs_inode *base_ni; @@ -3610,6 +3611,8 @@ static int ntfs_attr_update_meta(struct attr_record *a, struct ntfs_inode *ni, le16_to_cpu(a->data.non_resident.mapping_pairs_offset) == 8) && !(le32_to_cpu(m->bytes_allocated) - le32_to_cpu(m->bytes_in_use))) { + if (defer_attrlist) + return -ENOSPC; if (!NInoAttrList(base_ni)) { err = ntfs_inode_add_attrlist(base_ni); if (err) @@ -3623,7 +3626,7 @@ static int ntfs_attr_update_meta(struct attr_record *a, struct ntfs_inode *ni, goto out; } - err = ntfs_attrlist_update(base_ni); + err = ntfs_attrlist_update_locked(base_ni, locked_ni); if (err) goto out; err = -EAGAIN; @@ -3703,6 +3706,8 @@ out: * ntfs_attr_update_mapping_pairs - update mapping pairs for ntfs attribute * @ni: non-resident ntfs inode for which we need update * @from_vcn: update runlist starting this VCN + * @locked_ni: inode whose runlist write lock is already held + * @defer_attrlist: return -ENOSPC instead of updating an attribute list * * Build mapping pairs from @na->rl and write them to the disk. Also, this * function updates sparse bit, allocated and compressed size (allocates/frees @@ -3712,7 +3717,10 @@ out: * call to this function. Vice-versa @na->compressed_size will be calculated and * set to correct value during this function. */ -int ntfs_attr_update_mapping_pairs(struct ntfs_inode *ni, s64 from_vcn) +static int __ntfs_attr_update_mapping_pairs(struct ntfs_inode *ni, + s64 from_vcn, + struct ntfs_inode *locked_ni, + bool defer_attrlist) { struct ntfs_attr_search_ctx *ctx; struct ntfs_inode *base_ni; @@ -3804,7 +3812,8 @@ retry: continue; } - err = ntfs_attr_update_meta(a, ni, m, ctx); + err = ntfs_attr_update_meta(a, ni, m, ctx, locked_ni, + defer_attrlist); if (err < 0) { if (err == -EAGAIN) { ntfs_attr_put_search_ctx(ctx); @@ -3844,18 +3853,28 @@ retry: */ if (ni->type == AT_ATTRIBUTE_LIST) { ntfs_attr_put_search_ctx(ctx); - if (ntfs_inode_free_space(base_ni, mp_size - - cur_max_mp_size)) { - ntfs_debug("Attribute list is too big. Defragment the volume\n"); - return -ENOSPC; + ctx = NULL; + if (locked_ni == ni || defer_attrlist) { + err = -ENOSPC; + goto put_err_out; } - if (ntfs_attrlist_update(base_ni)) - return -EIO; + err = ntfs_inode_free_space(base_ni, mp_size - + cur_max_mp_size); + if (err) + return err; + err = ntfs_attrlist_update_locked( + base_ni, locked_ni); + if (err) + return err; goto retry; } /* Add attribute list if it isn't present, and retry. */ if (!NInoAttrList(base_ni)) { + if (defer_attrlist) { + err = -ENOSPC; + goto put_err_out; + } ntfs_attr_put_search_ctx(ctx); if (ntfs_inode_add_attrlist(base_ni)) { ntfs_error(sb, "Can not add attrlist"); @@ -3883,13 +3902,21 @@ retry: } } + if (defer_attrlist && + (ctx->ntfs_ino->nr_extents == -1 || + NInoAttrList(ctx->ntfs_ino)) && + ctx->attr->type != AT_ATTRIBUTE_LIST) { + err = -ENOSPC; + goto put_err_out; + } + /* Update lowest vcn. */ a->data.non_resident.lowest_vcn = cpu_to_le64(stop_vcn); mark_mft_record_dirty(ctx->ntfs_ino); if ((ctx->ntfs_ino->nr_extents == -1 || NInoAttrList(ctx->ntfs_ino)) && ctx->attr->type != AT_ATTRIBUTE_LIST) { ctx->al_entry->lowest_vcn = cpu_to_le64(stop_vcn); - err = ntfs_attrlist_update(base_ni); + err = ntfs_attrlist_update_locked(base_ni, locked_ni); if (err) goto put_err_out; } @@ -3976,7 +4003,10 @@ retry: unsigned int de_cnt = 0; /* Allocate new mft record. */ - err = ntfs_mft_record_alloc(ni->vol, 0, &ext_ni, base_ni, NULL); + err = ntfs_mft_record_alloc(ni->vol, 0, &ext_ni, base_ni, NULL, + base_ni->mft_no == FILE_MFT && + ni->type == AT_DATA && + ni->name == AT_UNNAMED ? stop_vcn : -1); if (err) { ntfs_error(sb, "Failed to allocate extent record"); goto put_err_out; @@ -4061,6 +4091,19 @@ put_err_out: return err; } +int ntfs_attr_update_mapping_pairs_locked(struct ntfs_inode *ni, + s64 from_vcn, + struct ntfs_inode *locked_ni) +{ + return __ntfs_attr_update_mapping_pairs(ni, from_vcn, locked_ni, + false); +} + +int ntfs_attr_update_mapping_pairs(struct ntfs_inode *ni, s64 from_vcn) +{ + return ntfs_attr_update_mapping_pairs_locked(ni, from_vcn, NULL); +} + /* * ntfs_attr_make_resident - convert a non-resident to a resident attribute * @ni: open ntfs attribute to make resident @@ -4194,7 +4237,9 @@ static int ntfs_attr_make_resident(struct ntfs_inode *ni, struct ntfs_attr_searc * * Reduce the size of a non-resident, open ntfs attribute @na to @newsize bytes. */ -static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsize) +static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, + const s64 newsize, + struct ntfs_inode *locked_ni) { struct ntfs_volume *vol; struct ntfs_attr_search_ctx *ctx; @@ -4202,6 +4247,7 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz s64 nr_freed_clusters; int err; struct ntfs_inode *base_ni; + bool runlist_locked = locked_ni == ni; ntfs_debug("Inode 0x%llx attr 0x%x new size %lld\n", (unsigned long long)ni->mft_no, ni->type, (long long)newsize); @@ -4247,18 +4293,24 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz * clusters if there is a change. */ if (ntfs_bytes_to_cluster(vol, ni->allocated_size) != first_free_vcn) { - struct ntfs_attr_search_ctx *ctx; + /* + * ntfs_cluster_free() and ntfs_rl_truncate_nolock() + * both require this lock. + */ + if (!runlist_locked) + down_write(&ni->runlist.lock); err = ntfs_attr_map_whole_runlist(ni); if (err) { ntfs_debug("Eeek! ntfs_attr_map_whole_runlist failed.\n"); - return err; + goto unlock_runlist; } ctx = ntfs_attr_get_search_ctx(ni, NULL); if (!ctx) { ntfs_error(vol->sb, "%s: Failed to get search context", __func__); - return -ENOMEM; + err = -ENOMEM; + goto unlock_runlist; } /* Deallocate all clusters starting with the first free one. */ @@ -4266,7 +4318,8 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz if (nr_freed_clusters < 0) { ntfs_debug("Eeek! Freeing of clusters failed. Aborting...\n"); ntfs_attr_put_search_ctx(ctx); - return (int)nr_freed_clusters; + err = (int)nr_freed_clusters; + goto unlock_runlist; } ntfs_attr_put_search_ctx(ctx); @@ -4279,7 +4332,8 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz kvfree(ni->runlist.rl); ni->runlist.rl = NULL; ntfs_error(vol->sb, "Eeek! Run list truncation failed.\n"); - return -EIO; + err = -EIO; + goto unlock_runlist; } /* Prepare to mapping pairs update. */ @@ -4295,11 +4349,13 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz VFS_I(base_ni)->i_blocks = ni->allocated_size >> 9; /* Write mapping pairs for new runlist. */ - err = ntfs_attr_update_mapping_pairs(ni, 0 /*first_free_vcn*/); + err = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); if (err) { ntfs_debug("Eeek! Mapping pairs update failed. Leaving inconstant metadata. Run chkdsk.\n"); - return err; + goto unlock_runlist; } + if (!runlist_locked) + up_write(&ni->runlist.lock); } /* Get the first attribute record. */ @@ -4341,7 +4397,11 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz /* If the attribute now has zero size, make it resident. */ if (!newsize && !NInoEncrypted(ni) && !NInoCompressed(ni)) { + if (!runlist_locked) + down_write(&ni->runlist.lock); err = ntfs_attr_make_resident(ni, ctx); + if (!runlist_locked) + up_write(&ni->runlist.lock); if (err) { /* If couldn't make resident, just continue. */ if (err != -EPERM) @@ -4358,6 +4418,11 @@ static int ntfs_non_resident_attr_shrink(struct ntfs_inode *ni, const s64 newsiz put_err_out: ntfs_attr_put_search_ctx(ctx); return err; + +unlock_runlist: + if (!runlist_locked) + up_write(&ni->runlist.lock); + return err; } /* @@ -4366,13 +4431,14 @@ put_err_out: * @prealloc_size: preallocation size (in bytes) to which to expand the attribute * @newsize: new size (in bytes) to which to expand the attribute * @holes: how to create a hole if expanding - * @need_lock: whether mrec lock is needed or not + * @locked_ni: inode whose runlist lock is already held * * Expand the size of a non-resident, open ntfs attribute @na to @newsize bytes, * by allocating new clusters. */ static int ntfs_non_resident_attr_expand(struct ntfs_inode *ni, const s64 newsize, - const s64 prealloc_size, unsigned int holes, bool need_lock) + const s64 prealloc_size, unsigned int holes, + struct ntfs_inode *locked_ni) { s64 lcn_seek_from; s64 first_free_vcn; @@ -4519,13 +4585,39 @@ static int ntfs_non_resident_attr_expand(struct ntfs_inode *ni, const s64 newsiz ntfs_bytes_to_cluster(vol, ni->allocated_size), first_free_vcn - ntfs_bytes_to_cluster(vol, ni->allocated_size), - lcn_seek_from, DATA_ZONE, false, false, false); + lcn_seek_from, DATA_ZONE, false, + ni->type == AT_ATTRIBUTE_LIST, false); if (IS_ERR(rl)) { ntfs_debug("Cluster allocation failed (%lld)", (long long)first_free_vcn - ntfs_bytes_to_cluster(vol, ni->allocated_size)); return PTR_ERR(rl); } + /* + * A contiguous ATTRIBUTE_LIST allocation keeps its mapping + * pairs small enough to fit in the base MFT record. The + * allocator can return a short run when contiguity was + * requested, so discard it and retry normally if necessary. + */ + if (ni->type == AT_ATTRIBUTE_LIST && + (rl->vcn != ntfs_bytes_to_cluster(vol, + ni->allocated_size) || + rl->length != first_free_vcn - + ntfs_bytes_to_cluster(vol, ni->allocated_size) || + rl[1].length)) { + ntfs_cluster_free_from_rl(vol, rl); + kvfree(rl); + rl = ntfs_cluster_alloc(vol, + ntfs_bytes_to_cluster(vol, + ni->allocated_size), + first_free_vcn - + ntfs_bytes_to_cluster(vol, + ni->allocated_size), + lcn_seek_from, DATA_ZONE, false, + false, false); + if (IS_ERR(rl)) + return PTR_ERR(rl); + } } if (!NInoCompressed(ni)) { @@ -4544,7 +4636,8 @@ static int ntfs_non_resident_attr_expand(struct ntfs_inode *ni, const s64 newsiz /* Prepare to mapping pairs update. */ ni->allocated_size = ntfs_cluster_to_bytes(vol, first_free_vcn); - err = ntfs_attr_update_mapping_pairs(ni, 0); + err = ntfs_attr_update_mapping_pairs_locked( + ni, 0, locked_ni); if (err) { ntfs_debug("Mapping pairs update failed"); goto rollback; @@ -4588,11 +4681,11 @@ rollback: ntfs_debug("Leaking clusters"); /* Now, truncate the runlist itself. */ - if (need_lock) + if (ni != locked_ni) down_write(&ni->runlist.lock); err2 = ntfs_rl_truncate_nolock(vol, &ni->runlist, ntfs_bytes_to_cluster(vol, org_alloc_size)); - if (need_lock) + if (ni != locked_ni) up_write(&ni->runlist.lock); if (err2) { /* @@ -4606,11 +4699,11 @@ rollback: /* Prepare to mapping pairs update. */ ni->allocated_size = org_alloc_size; /* Restore mapping pairs. */ - if (need_lock) + if (ni != locked_ni) down_read(&ni->runlist.lock); - if (ntfs_attr_update_mapping_pairs(ni, 0)) + if (__ntfs_attr_update_mapping_pairs(ni, 0, locked_ni, true)) ntfs_error(sb, "Failed to restore old mapping pairs"); - if (need_lock) + if (ni != locked_ni) up_read(&ni->runlist.lock); if (NInoSparse(ni) || NInoCompressed(ni)) { @@ -4715,7 +4808,8 @@ attr_resize_again: mark_mft_record_dirty(ctx->ntfs_ino); ntfs_attr_put_search_ctx(ctx); /* Resize non-resident attribute */ - return ntfs_non_resident_attr_expand(attr_ni, newsize, prealloc_size, holes, true); + return ntfs_non_resident_attr_expand( + attr_ni, newsize, prealloc_size, holes, NULL); } else if (err != -ENOSPC && err != -EPERM) { ntfs_error(sb, "Failed to make attribute non-resident"); goto put_err_out; @@ -4836,7 +4930,7 @@ attr_resize_again: } /* Allocate new mft record. */ - err = ntfs_mft_record_alloc(base_ni->vol, 0, &ext_ni, base_ni, NULL); + err = ntfs_mft_record_alloc(base_ni->vol, 0, &ext_ni, base_ni, NULL, -1); if (err) { ntfs_error(sb, "Couldn't allocate MFT record"); goto put_err_out; @@ -4890,13 +4984,14 @@ int __ntfs_attr_truncate_vfs(struct ntfs_inode *ni, const s64 newsize, if (NInoNonResident(ni)) { if (newsize > i_size) { down_write(&ni->runlist.lock); - err = ntfs_non_resident_attr_expand(ni, newsize, 0, - NVolDisableSparse(ni->vol) ? - HOLES_NO : HOLES_OK, - false); + err = ntfs_non_resident_attr_expand( + ni, newsize, 0, + NVolDisableSparse(ni->vol) ? + HOLES_NO : HOLES_OK, ni); up_write(&ni->runlist.lock); } else - err = ntfs_non_resident_attr_shrink(ni, newsize); + err = ntfs_non_resident_attr_shrink( + ni, newsize, NULL); } else err = ntfs_resident_attr_resize(ni, newsize, 0, NVolDisableSparse(ni->vol) ? @@ -4905,7 +5000,9 @@ int __ntfs_attr_truncate_vfs(struct ntfs_inode *ni, const s64 newsize, return err; } -int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, const s64 prealloc_size) +int ntfs_attr_expand_locked(struct ntfs_inode *ni, const s64 newsize, + const s64 prealloc_size, + struct ntfs_inode *locked_ni) { int err = 0; @@ -4918,7 +5015,8 @@ int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, const s64 preallo ntfs_debug("Entering for inode 0x%llx, attr 0x%x, size %lld\n", (unsigned long long)ni->mft_no, ni->type, newsize); - if (ni->data_size == newsize) { + if (ni->data_size == newsize && + (!prealloc_size || prealloc_size <= ni->allocated_size)) { ntfs_debug("Size is already ok\n"); return 0; } @@ -4933,10 +5031,11 @@ int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, const s64 preallo } if (NInoNonResident(ni)) { - if (newsize > ni->data_size) - err = ntfs_non_resident_attr_expand(ni, newsize, prealloc_size, - NVolDisableSparse(ni->vol) ? - HOLES_NO : HOLES_OK, true); + if (newsize > ni->data_size || prealloc_size > ni->allocated_size) + err = ntfs_non_resident_attr_expand( + ni, newsize, prealloc_size, + NVolDisableSparse(ni->vol) ? + HOLES_NO : HOLES_OK, locked_ni); } else err = ntfs_resident_attr_resize(ni, newsize, prealloc_size, NVolDisableSparse(ni->vol) ? @@ -4947,6 +5046,12 @@ int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, const s64 preallo return err; } +int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, + const s64 prealloc_size) +{ + return ntfs_attr_expand_locked(ni, newsize, prealloc_size, NULL); +} + /* * ntfs_attr_truncate_i - resize an ntfs attribute * @ni: open ntfs inode to resize @@ -4959,7 +5064,9 @@ int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, const s64 preallo * newly allocated space is marked as not initialised and no real allocation * on disk is performed. */ -int ntfs_attr_truncate_i(struct ntfs_inode *ni, const s64 newsize, unsigned int holes) +int ntfs_attr_truncate_i_locked(struct ntfs_inode *ni, const s64 newsize, + unsigned int holes, + struct ntfs_inode *locked_ni) { int err; @@ -4993,15 +5100,23 @@ int ntfs_attr_truncate_i(struct ntfs_inode *ni, const s64 newsize, unsigned int if (NInoNonResident(ni)) { if (newsize > ni->data_size) - err = ntfs_non_resident_attr_expand(ni, newsize, 0, holes, true); + err = ntfs_non_resident_attr_expand( + ni, newsize, 0, holes, locked_ni); else - err = ntfs_non_resident_attr_shrink(ni, newsize); + err = ntfs_non_resident_attr_shrink( + ni, newsize, locked_ni); } else err = ntfs_resident_attr_resize(ni, newsize, 0, holes); ntfs_debug("Return status %d\n", err); return err; } +int ntfs_attr_truncate_i(struct ntfs_inode *ni, const s64 newsize, + unsigned int holes) +{ + return ntfs_attr_truncate_i_locked(ni, newsize, holes, NULL); +} + /* * Resize an attribute, creating a hole if relevant */ @@ -5019,10 +5134,11 @@ int ntfs_attr_map_cluster(struct ntfs_inode *ni, s64 vcn_start, s64 *lcn_start, struct ntfs_volume *vol = ni->vol; struct ntfs_attr_search_ctx *ctx; struct runlist_element *rl, *rlc; + struct runlist_element *old_rl = NULL; s64 vcn = vcn_start, lcn, clu_count; s64 lcn_seek_from = -1; int err = 0; - size_t new_rl_count; + size_t new_rl_count, old_rl_count; err = ntfs_attr_map_whole_runlist(ni); if (err) @@ -5115,6 +5231,19 @@ int ntfs_attr_map_cluster(struct ntfs_inode *ni, s64 vcn_start, s64 *lcn_start, WARN_ON(rlc->vcn != vcn); lcn = rlc->lcn; clu_count = rlc->length; + old_rl_count = ni->runlist.count; + old_rl = kmemdup(ni->runlist.rl, + old_rl_count * sizeof(*old_rl), GFP_NOFS); + if (!old_rl) { + err = -ENOMEM; + if (ntfs_cluster_free_from_rl(vol, rlc)) { + ntfs_error(vol->sb, + "Failed to free cluster allocation after runlist backup failure."); + NVolSetErrors(vol); + } + kvfree(rlc); + goto out; + } rl = ntfs_runlists_merge(&ni->runlist, rlc, 0, &new_rl_count); if (IS_ERR(rl)) { @@ -5138,15 +5267,32 @@ int ntfs_attr_map_cluster(struct ntfs_inode *ni, s64 vcn_start, s64 *lcn_start, if (update_mp) { ntfs_attr_reinit_search_ctx(ctx); - err = ntfs_attr_update_mapping_pairs(ni, 0); + err = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); if (err) { int err2; err2 = ntfs_cluster_free(ni, vcn, clu_count, ctx); - if (err2 < 0) + if (err2 < 0 || err2 != clu_count) { ntfs_error(vol->sb, - "Failed to free cluster allocation. Leaving inconstant metadata.\n"); - goto out; + "Failed to free cluster allocation. Leaving inconsistent metadata.\n"); + NVolSetErrors(vol); + goto out; + } + + /* + * Restore the runlist before repairing the on-disk + * mapping pairs. + */ + kvfree(ni->runlist.rl); + ni->runlist.rl = old_rl; + ni->runlist.count = old_rl_count; + old_rl = NULL; + if (ntfs_attr_update_mapping_pairs_locked( + ni, 0, ni)) { + ntfs_error(vol->sb, + "Failed to restore mapping pairs after allocation rollback.\n"); + NVolSetErrors(vol); + } } } else { VFS_I(ni)->i_blocks += clu_count << (vol->cluster_size_bits - 9); @@ -5158,6 +5304,7 @@ int ntfs_attr_map_cluster(struct ntfs_inode *ni, s64 vcn_start, s64 *lcn_start, *lcn_count = clu_count; *balloc = true; out: + kvfree(old_rl); ntfs_attr_put_search_ctx(ctx); return err; } @@ -5401,7 +5548,7 @@ int ntfs_non_resident_attr_insert_range(struct ntfs_inode *ni, s64 start_vcn, s6 ni->data_size += ntfs_cluster_to_bytes(vol, len); if (ntfs_cluster_to_bytes(vol, start_vcn) < ni->initialized_size) ni->initialized_size += ntfs_cluster_to_bytes(vol, len); - ret = ntfs_attr_update_mapping_pairs(ni, 0); + ret = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); up_write(&ni->runlist.lock); if (ret) return ret; @@ -5486,7 +5633,7 @@ int ntfs_non_resident_attr_collapse_range(struct ntfs_inode *ni, s64 start_vcn, } if (ni->allocated_size > 0) { - ret = ntfs_attr_update_mapping_pairs(ni, 0); + ret = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); if (ret) { up_write(&ni->runlist.lock); goto out_rl; @@ -5564,7 +5711,7 @@ int ntfs_non_resident_attr_punch_hole(struct ntfs_inode *ni, s64 start_vcn, s64 ni->runlist.rl = rl; ni->runlist.count = new_rl_count; - ret = ntfs_attr_update_mapping_pairs(ni, 0); + ret = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); up_write(&ni->runlist.lock); if (ret) { kvfree(punch_rl); @@ -5740,7 +5887,7 @@ int ntfs_attr_fallocate(struct ntfs_inode *ni, loff_t start, loff_t byte_len, bo if (NInoRunlistDirty(ni)) { mutex_lock_nested(&ni->mrec_lock, NTFS_INODE_MUTEX_NORMAL); down_write(&ni->runlist.lock); - err = ntfs_attr_update_mapping_pairs(ni, 0); + err = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); if (err) ntfs_error(ni->vol->sb, "Updating mapping pairs failed"); else diff --git a/fs/ntfs/attrib.h b/fs/ntfs/attrib.h index e2224fbfaabe..6b4fa9f57640 100644 --- a/fs/ntfs/attrib.h +++ b/fs/ntfs/attrib.h @@ -112,7 +112,13 @@ int ntfs_non_resident_attr_punch_hole(struct ntfs_inode *ni, s64 start_vcn, s64 int __ntfs_attr_truncate_vfs(struct ntfs_inode *ni, const s64 newsize, const s64 i_size); int ntfs_attr_expand(struct ntfs_inode *ni, const s64 newsize, const s64 prealloc_size); +int ntfs_attr_expand_locked(struct ntfs_inode *ni, const s64 newsize, + const s64 prealloc_size, + struct ntfs_inode *locked_ni); int ntfs_attr_truncate_i(struct ntfs_inode *ni, const s64 newsize, unsigned int holes); +int ntfs_attr_truncate_i_locked(struct ntfs_inode *ni, const s64 newsize, + unsigned int holes, + struct ntfs_inode *locked_ni); int ntfs_attr_truncate(struct ntfs_inode *ni, const s64 newsize); int ntfs_attr_rm(struct ntfs_inode *ni); int ntfs_attr_exist(struct ntfs_inode *ni, const __le32 type, __le16 *name, @@ -133,6 +139,9 @@ int ntfs_resident_attr_record_add(struct ntfs_inode *ni, __le32 type, __le16 *name, u8 name_len, u8 *val, u32 size, __le16 flags); int ntfs_attr_update_mapping_pairs(struct ntfs_inode *ni, s64 from_vcn); +int ntfs_attr_update_mapping_pairs_locked(struct ntfs_inode *ni, + s64 from_vcn, + struct ntfs_inode *locked_ni); struct runlist_element *ntfs_attr_vcn_to_rl(struct ntfs_inode *ni, s64 vcn, s64 *lcn); /* diff --git a/fs/ntfs/attrlist.c b/fs/ntfs/attrlist.c index be3086d34338..bb191953dcb1 100644 --- a/fs/ntfs/attrlist.c +++ b/fs/ntfs/attrlist.c @@ -12,6 +12,9 @@ #include "mft.h" #include "attrib.h" #include "attrlist.h" +#include "lcnalloc.h" + +#define NTFS_MAX_ATTR_LIST_SIZE (256 * 1024) /* * ntfs_attrlist_need - check whether inode need attribute list @@ -51,11 +54,155 @@ int ntfs_attrlist_need(struct ntfs_inode *ni) return 0; } -int ntfs_attrlist_update(struct ntfs_inode *base_ni) +/* + * Repack the $MFT/$ATTRIBUTE_LIST data into one run. + * + * The mapping pairs for an $ATTRIBUTE_LIST must remain in the base MFT + * record. Once that record has no room left, extending a fragmented list + * can require one more mapping-pairs byte than the record can hold. There + * is no attribute that can legally be moved out in that state: $STANDARD_ + * INFORMATION, $ATTRIBUTE_LIST, and the first $MFT/$DATA extent all have to + * stay in the base record. Move the list data to one contiguous run. The + * caller supplies the minimum allocation size so a recovery can use the + * smallest useful run while normal updates can still request the maximum + * legal list size as a reserve. + */ +static int ntfs_attrlist_repack(struct inode *attr_vi, + struct ntfs_inode *attr_ni, s64 min_alloc_size, + struct ntfs_inode *locked_ni) +{ + struct ntfs_volume *vol = attr_ni->vol; + struct runlist_element *old_rl, *new_rl; + u8 *data = NULL; + s64 data_size, alloc_size, nr_clusters, written; + s64 old_alloc_size; + size_t old_rl_count, new_rl_count; + unsigned long flags; + int err, restore_err; + if (attr_ni->mft_no != FILE_MFT || !NInoNonResident(attr_ni) || + min_alloc_size < 0) + return -EINVAL; + /* The buffered I/O below can reacquire the attribute runlist lock. */ + if (attr_ni == locked_ni) + return -ENOSPC; + + err = ntfs_attr_map_whole_runlist(attr_ni); + if (err) + return err; + + data_size = attr_ni->data_size; + if (data_size < 0) + return -EIO; + + if (data_size) { + data = kvmalloc(data_size, GFP_NOFS); + if (!data) + return -ENOMEM; + + written = ntfs_inode_attr_pread(attr_vi, 0, data_size, data); + if (written != data_size) { + err = written < 0 ? (int)written : -EIO; + goto out_free_data; + } + } + + old_alloc_size = attr_ni->allocated_size; + alloc_size = max_t(s64, old_alloc_size, min_alloc_size); + nr_clusters = ntfs_bytes_to_cluster(vol, + alloc_size + vol->cluster_size - 1); + if (nr_clusters <= 0) { + err = -EFBIG; + goto out_free_data; + } + + /* A single run keeps the mapping pairs at the minimum size. */ + new_rl = ntfs_cluster_alloc(vol, 0, nr_clusters, -1, DATA_ZONE, + true, true, false); + if (IS_ERR(new_rl)) { + err = PTR_ERR(new_rl); + goto out_free_data; + } + + new_rl_count = 0; + if (new_rl->vcn == 0 && new_rl->length == nr_clusters && + !new_rl[1].length) + new_rl_count = 2; + + if (new_rl_count != 2) { + ntfs_cluster_free_from_rl(vol, new_rl); + kvfree(new_rl); + err = -ENOSPC; + goto out_free_data; + } + old_rl = attr_ni->runlist.rl; + old_rl_count = attr_ni->runlist.count; + down_write(&attr_ni->runlist.lock); + attr_ni->runlist.rl = new_rl; + attr_ni->runlist.count = new_rl_count; + up_write(&attr_ni->runlist.lock); + + write_lock_irqsave(&attr_ni->size_lock, flags); + attr_ni->allocated_size = ntfs_cluster_to_bytes(vol, nr_clusters); + write_unlock_irqrestore(&attr_ni->size_lock, flags); + + /* Populate the replacement extent before publishing its mapping pairs. */ + if (data_size) { + written = ntfs_inode_attr_pwrite(attr_vi, 0, data_size, data, true); + if (written != data_size) { + err = written < 0 ? (int)written : -EIO; + goto restore_old_runlist; + } + } + + err = ntfs_attr_update_mapping_pairs_locked(attr_ni, 0, locked_ni); + if (err) + goto restore_old_runlist; + + /* The new mapping is now authoritative; release the old data runs. */ + if (ntfs_cluster_free_from_rl(vol, old_rl)) { + ntfs_error(vol->sb, + "Failed to free old ATTRIBUTE_LIST extent: inode %#llx", + (long long)attr_ni->mft_no); + NVolSetErrors(vol); + } + kvfree(old_rl); + kvfree(data); + return 0; + +restore_old_runlist: + down_write(&attr_ni->runlist.lock); + attr_ni->runlist.rl = old_rl; + attr_ni->runlist.count = old_rl_count; + up_write(&attr_ni->runlist.lock); + + write_lock_irqsave(&attr_ni->size_lock, flags); + attr_ni->allocated_size = old_alloc_size; + write_unlock_irqrestore(&attr_ni->size_lock, flags); + + restore_err = ntfs_attr_update_mapping_pairs_locked( + attr_ni, 0, locked_ni); + if (restore_err) { + ntfs_error(vol->sb, "Failed to restore ATTRIBUTE_LIST mapping pairs (%d)", + restore_err); + NVolSetErrors(vol); + } + + ntfs_cluster_free_from_rl(vol, new_rl); + kvfree(new_rl); + err = err ? err : restore_err; + +out_free_data: + kvfree(data); + return err; +} + +int ntfs_attrlist_update_locked(struct ntfs_inode *base_ni, + struct ntfs_inode *locked_ni) { struct inode *attr_vi; struct ntfs_inode *attr_ni; - int err; + s64 written; + int err, retry_err; /* * generic_shutdown_super() clears SB_ACTIVE before evicting cached @@ -72,23 +219,66 @@ int ntfs_attrlist_update(struct ntfs_inode *base_ni) return err; } attr_ni = NTFS_I(attr_vi); + /* Truncation and page-cache writes can reacquire this runlist lock. */ + if (attr_ni == locked_ni) { + iput(attr_vi); + return -ENOSPC; + } - err = ntfs_attr_truncate_i(attr_ni, base_ni->attr_list_size, HOLES_NO); - if (err == -ENOSPC && attr_ni->mft_no == FILE_MFT) { - err = ntfs_attr_truncate(attr_ni, 0); - if (err || ntfs_attr_truncate_i(attr_ni, base_ni->attr_list_size, HOLES_NO) != 0) { + err = ntfs_attr_truncate_i_locked( + attr_ni, base_ni->attr_list_size, HOLES_NO, locked_ni); + if (err == -ENOSPC && attr_ni->mft_no == FILE_MFT && + NInoNonResident(attr_ni)) { + retry_err = ntfs_attrlist_repack(attr_vi, attr_ni, + base_ni->attr_list_size, locked_ni); + if (retry_err) { + ntfs_error(base_ni->vol->sb, "Failed to repack attribute list"); iput(attr_vi); + return retry_err; + } + + retry_err = ntfs_attr_truncate_i_locked( + attr_ni, base_ni->attr_list_size, + HOLES_NO, locked_ni); + if (retry_err) { ntfs_error(base_ni->vol->sb, - "Failed to truncate attribute list of inode %#llx", - (long long)base_ni->mft_no); - return -EIO; + "Failed to resize attribute list after repack"); + iput(attr_vi); + return retry_err; } } else if (err) { iput(attr_vi); ntfs_error(base_ni->vol->sb, "Failed to truncate attribute list of inode %#llx", (long long)base_ni->mft_no); - return -EIO; + return err; + } + + /* + * Reserve the maximum legal list size while the MFT metadata area is + * still easy to allocate contiguously. This prevents a later list entry + * from needing another mapping-pairs byte in the full base MFT record. + * Failure to obtain the optional reserve must not reject the current + * metadata update; the repack retry above remains available if needed. + */ + if (base_ni->mft_no == FILE_MFT && NInoNonResident(attr_ni) && + attr_ni->allocated_size < NTFS_MAX_ATTR_LIST_SIZE) { + retry_err = ntfs_attr_expand_locked( + attr_ni, base_ni->attr_list_size, + NTFS_MAX_ATTR_LIST_SIZE, locked_ni); + if (retry_err == -ENOSPC) { + retry_err = ntfs_attrlist_repack( + attr_vi, attr_ni, + NTFS_MAX_ATTR_LIST_SIZE, locked_ni); + if (retry_err == -ENOSPC) + retry_err = 0; + } + if (retry_err) { + ntfs_error(base_ni->vol->sb, + "Failed to reserve attribute list space"); + iput(attr_vi); + return retry_err; + } } i_size_write(attr_vi, base_ni->attr_list_size); @@ -96,14 +286,15 @@ int ntfs_attrlist_update(struct ntfs_inode *base_ni) if (NInoNonResident(attr_ni) && !NInoAttrListNonResident(base_ni)) NInoSetAttrListNonResident(base_ni); - if (ntfs_inode_attr_pwrite(attr_vi, 0, base_ni->attr_list_size, - base_ni->attr_list, false) != - base_ni->attr_list_size) { + written = ntfs_inode_attr_pwrite(attr_vi, 0, base_ni->attr_list_size, + base_ni->attr_list, false); + if (written != base_ni->attr_list_size) { + err = written < 0 ? (int)written : -EIO; iput(attr_vi); ntfs_error(base_ni->vol->sb, "Failed to write attribute list of inode %#llx", (long long)base_ni->mft_no); - return -EIO; + return err; } NInoSetAttrListDirty(base_ni); @@ -111,6 +302,11 @@ int ntfs_attrlist_update(struct ntfs_inode *base_ni) return 0; } +int ntfs_attrlist_update(struct ntfs_inode *base_ni) +{ + return ntfs_attrlist_update_locked(base_ni, NULL); +} + /* * ntfs_attrlist_entry_add - add an attribute list attribute entry * @ni: opened ntfs inode, which contains that attribute diff --git a/fs/ntfs/attrlist.h b/fs/ntfs/attrlist.h index 1892a3934d3a..10cc2cc8e208 100644 --- a/fs/ntfs/attrlist.h +++ b/fs/ntfs/attrlist.h @@ -16,5 +16,7 @@ int ntfs_attrlist_need(struct ntfs_inode *ni); int ntfs_attrlist_entry_add(struct ntfs_inode *ni, struct attr_record *attr); int ntfs_attrlist_entry_rm(struct ntfs_attr_search_ctx *ctx); int ntfs_attrlist_update(struct ntfs_inode *base_ni); +int ntfs_attrlist_update_locked(struct ntfs_inode *base_ni, + struct ntfs_inode *locked_ni); #endif /* defined _NTFS_ATTRLIST_H */ diff --git a/fs/ntfs/compress.c b/fs/ntfs/compress.c index 99a3ea2b5c55..075b57fc1de6 100644 --- a/fs/ntfs/compress.c +++ b/fs/ntfs/compress.c @@ -1450,7 +1450,7 @@ static int ntfs_write_cb(struct ntfs_inode *ni, loff_t pos, struct page **pages, ni->runlist.rl = rl; rlc = NULL; - err = ntfs_attr_update_mapping_pairs(ni, 0); + err = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); up_write(&ni->runlist.lock); if (err) err = -EIO; diff --git a/fs/ntfs/file.c b/fs/ntfs/file.c index 8164326b7812..007d1614b9ac 100644 --- a/fs/ntfs/file.c +++ b/fs/ntfs/file.c @@ -111,7 +111,8 @@ static int ntfs_trim_prealloc(struct inode *vi) ntfs_error(vol->sb, "Preallocated block rollback failed"); } else { ni->allocated_size = ntfs_cluster_to_bytes(vol, vcn_tr); - err = ntfs_attr_update_mapping_pairs(ni, 0); + err = ntfs_attr_update_mapping_pairs_locked( + ni, 0, ni); if (err) ntfs_error(vol->sb, "Failed to rollback mapping pairs for prealloc"); diff --git a/fs/ntfs/inode.c b/fs/ntfs/inode.c index 5aedc045f65a..a777de8a80c7 100644 --- a/fs/ntfs/inode.c +++ b/fs/ntfs/inode.c @@ -170,17 +170,19 @@ struct inode *ntfs_iget(struct super_block *sb, u64 mft_no) /* If this is a freshly allocated inode, need to read it now. */ if (inode_state_read_once(vi) & I_NEW) { err = ntfs_read_locked_inode(vi); - unlock_new_inode(vi); + if (err) { + remove_inode_hash(vi); + discard_new_inode(vi); + } else + unlock_new_inode(vi); } /* * There is no point in keeping bad inodes around. This also * simplifies things in that we never need to check for bad inodes * elsewhere. */ - if (unlikely(err)) { - iput(vi); + if (unlikely(err)) vi = ERR_PTR(err); - } return vi; } @@ -231,17 +233,19 @@ struct inode *ntfs_attr_iget(struct inode *base_vi, __le32 type, /* If this is a freshly allocated inode, need to read it now. */ if (inode_state_read_once(vi) & I_NEW) { err = ntfs_read_locked_attr_inode(base_vi, vi); - unlock_new_inode(vi); + if (err) { + remove_inode_hash(vi); + discard_new_inode(vi); + } else + unlock_new_inode(vi); } /* * There is no point in keeping bad attribute inodes around. This also * simplifies things in that we never need to check for bad attribute * inodes elsewhere. */ - if (unlikely(err)) { - iput(vi); + if (unlikely(err)) vi = ERR_PTR(err); - } return vi; } @@ -286,17 +290,19 @@ struct inode *ntfs_index_iget(struct inode *base_vi, __le16 *name, /* If this is a freshly allocated inode, need to read it now. */ if (inode_state_read_once(vi) & I_NEW) { err = ntfs_read_locked_index_inode(base_vi, vi); - unlock_new_inode(vi); + if (err) { + remove_inode_hash(vi); + discard_new_inode(vi); + } else + unlock_new_inode(vi); } /* * There is no point in keeping bad index inodes around. This also * simplifies things in that we never need to check for bad index * inodes elsewhere. */ - if (unlikely(err)) { - iput(vi); + if (unlikely(err)) vi = ERR_PTR(err); - } return vi; } @@ -1241,7 +1247,8 @@ unm_err_out: if (m) unmap_mft_record(ni); err_out: - if (err != -EOPNOTSUPP && err != -ENOMEM && vol_err == true) { + if (err != -EOPNOTSUPP && err != -ENOMEM && + err != -EINTR && err != -ERESTARTSYS && vol_err == true) { ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt inode 0x%llx as bad. Run chkdsk.", err, ni->mft_no); @@ -1467,12 +1474,13 @@ unm_err_out: ntfs_attr_put_search_ctx(ctx); unmap_mft_record(base_ni); err_out: - if (err != -ENOENT) + if (err != -ENOENT && err != -EINTR && err != -ERESTARTSYS) ntfs_error(vol->sb, "Failed with error code %i while reading attribute inode (mft_no 0x%llx, type 0x%x, name_len %i). Marking corrupt inode and base inode 0x%llx as bad. Run chkdsk.", err, ni->mft_no, ni->type, ni->name_len, base_ni->mft_no); - if (err != -ENOENT && err != -ENOMEM) + if (err != -ENOENT && err != -ENOMEM && + err != -EINTR && err != -ERESTARTSYS) NVolSetErrors(vol); return err; } @@ -1676,8 +1684,9 @@ static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi) /* Get the index bitmap attribute inode. */ bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len); if (IS_ERR(bvi)) { - ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); err = PTR_ERR(bvi); + if (err != -EINTR && err != -ERESTARTSYS) + ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); goto unm_err_out; } bni = NTFS_I(bvi); @@ -1721,10 +1730,12 @@ unm_err_out: if (m) unmap_mft_record(base_ni); err_out: - ntfs_error(vi->i_sb, - "Failed with error code %i while reading index inode (mft_no 0x%llx, name_len %i.", - err, ni->mft_no, ni->name_len); - if (err != -EOPNOTSUPP && err != -ENOMEM) + if (err != -EINTR && err != -ERESTARTSYS) + ntfs_error(vi->i_sb, + "Failed with error code %i while reading index inode (mft_no 0x%llx, name_len %i.", + err, ni->mft_no, ni->name_len); + if (err != -EOPNOTSUPP && err != -ENOMEM && + err != -EINTR && err != -ERESTARTSYS) NVolSetErrors(vol); return err; } @@ -2772,7 +2783,7 @@ int __ntfs_write_inode(struct inode *vi, int sync) if (NInoNonResident(ni) && NInoRunlistDirty(ni)) { down_write(&ni->runlist.lock); - err = ntfs_attr_update_mapping_pairs(ni, 0); + err = ntfs_attr_update_mapping_pairs_locked(ni, 0, ni); if (!err) NInoClearRunlistDirty(ni); up_write(&ni->runlist.lock); @@ -3713,7 +3724,7 @@ static s64 __ntfs_inode_non_resident_attr_pwrite(struct inode *vi, FGP_CREAT | FGP_LOCK, mapping_gfp_mask(mapping)); if (IS_ERR(folio)) { - ret = -ENOMEM; + ret = PTR_ERR(folio); break; } } else { @@ -3745,6 +3756,7 @@ static s64 __ntfs_inode_non_resident_attr_pwrite(struct inode *vi, u64 rl_length = 0; s64 vcn; struct runlist_element *rl; + int bio_err; lcn_count = max_t(s64, 1, ntfs_bytes_to_cluster(vol, attr_len)); vcn = ntfs_pidx_to_cluster(vol, folio->index); @@ -3787,8 +3799,15 @@ static s64 __ntfs_inode_non_resident_attr_pwrite(struct inode *vi, goto err_unlock_folio; } - submit_bio_wait(bio); + bio_err = submit_bio_wait(bio); bio_put(bio); + if (bio_err) { + ntfs_error(vi->i_sb, + "Synchronous attribute write failed (%d)", + bio_err); + ret = bio_err; + goto err_unlock_folio; + } vcn += rl_length; offset += length; } while (lcn_count != 0); diff --git a/fs/ntfs/mft.c b/fs/ntfs/mft.c index 98ab686a5ea2..4b7449495375 100644 --- a/fs/ntfs/mft.c +++ b/fs/ntfs/mft.c @@ -213,7 +213,8 @@ struct mft_record *map_mft_record(struct ntfs_inode *ni) return m; atomic_dec(&ni->count); - ntfs_error(ni->vol->sb, "Failed with error code %lu.", -PTR_ERR(m)); + if (PTR_ERR(m) != -EINTR && PTR_ERR(m) != -ERESTARTSYS) + ntfs_error(ni->vol->sb, "Failed with error code %lu.", -PTR_ERR(m)); return m; } @@ -462,7 +463,7 @@ int ntfs_sync_mft_mirror(struct ntfs_volume *vol, const u64 mft_no, { u8 *kmirr; struct folio *folio; - unsigned int folio_ofs, lcn_folio_off = 0; + unsigned int folio_ofs; int err = 0; struct bio *bio; @@ -492,15 +493,11 @@ int ntfs_sync_mft_mirror(struct ntfs_volume *vol, const u64 mft_no, memcpy(kmirr, m, vol->mft_record_size); kunmap_local(kmirr); - if (vol->cluster_size_bits > PAGE_SHIFT) { - lcn_folio_off = folio->index << PAGE_SHIFT; - lcn_folio_off &= vol->cluster_size_mask; - } - bio = bio_alloc(vol->sb->s_bdev, 1, REQ_OP_WRITE, GFP_NOIO); bio->bi_iter.bi_sector = ntfs_bytes_to_bio_sector(NTFS_CLU_TO_B(vol, vol->mftmirr_lcn) + - lcn_folio_off + folio_ofs); + ((u64)folio->index << PAGE_SHIFT) + + folio_ofs); if (bio_add_folio(bio, folio, vol->mft_record_size, folio_ofs)) err = submit_bio_wait(bio); @@ -841,12 +838,126 @@ static bool ntfs_may_write_mft_record(struct ntfs_volume *vol, const u64 mft_no, static const char *es = " Leaving inconsistent metadata. Unmount and run chkdsk."; -#define RESERVED_MFT_RECORDS 64 +#define FIRST_NORMAL_MFT_RECORD 24 +#define MFT_RECORD_RESERVE 4 /* - * ntfs_mft_bitmap_find_and_alloc_free_rec_nolock - see name + * Records 12-15 are marked in use by Windows but normally have no name + * and no links. Keep them as the last bootstrap option when a volume + * mounted without an in-memory tail reserve needs its first $MFT metadata + * extent. + */ +static bool mft_reserved_is_free(struct ntfs_volume *vol, + struct ntfs_inode *mft_ni, s64 mft_no) +{ + struct attr_record *a; + struct mft_record *m; + struct folio *folio; + void *mapped; + pgoff_t index = NTFS_MFT_NR_TO_PIDX(vol, mft_no); + unsigned int ofs = NTFS_MFT_NR_TO_POFS(vol, mft_no); + u32 attrs_offset, bytes_in_use; + bool available = false, have_std = false; + int i; + + for (i = 0; i < mft_ni->nr_extents; i++) { + if (mft_ni->ext.extent_ntfs_inos[i] && + mft_ni->ext.extent_ntfs_inos[i]->mft_no == mft_no) + return false; + } + m = kmalloc(vol->mft_record_size, GFP_NOFS); + if (!m) + return false; + + folio = read_mapping_folio(vol->mft_ino->i_mapping, index, NULL); + if (IS_ERR(folio)) + goto free_m; + + folio_lock(folio); + mapped = kmap_local_folio(folio, 0); + memcpy(m, (u8 *)mapped + ofs, vol->mft_record_size); + kunmap_local(mapped); + folio_unlock(folio); + folio_put(folio); + if (post_read_mst_fixup((struct ntfs_record *)m, vol->mft_record_size)) + goto free_m; + + if (!ntfs_is_mft_record(m->magic) || + !(m->flags & MFT_RECORD_IN_USE) || m->base_mft_record || + m->link_count) + goto out; + + attrs_offset = le16_to_cpu(m->attrs_offset); + bytes_in_use = le32_to_cpu(m->bytes_in_use); + if (attrs_offset > bytes_in_use || bytes_in_use > vol->mft_record_size || + bytes_in_use - attrs_offset < sizeof(a->type)) + goto out; + + for (a = (struct attr_record *)((u8 *)m + attrs_offset); + (u8 *)a + sizeof(a->type) <= (u8 *)m + bytes_in_use;) { + u32 len; + + if (a->type == AT_END) { + if ((u8 *)a + sizeof(a->type) + sizeof(a->length) > + (u8 *)m + bytes_in_use) + break; + /* Also accept a record emptied by an earlier bootstrap. */ + available = have_std || + (u8 *)a == (u8 *)m + attrs_offset; + break; + } + if (a->type == AT_FILE_NAME) + break; + len = le32_to_cpu(a->length); + if (len < offsetof(struct attr_record, data) || + (u8 *)a + len > (u8 *)m + bytes_in_use) + break; + if (a->type == AT_STANDARD_INFORMATION) { + u32 value_len, value_ofs; + + if (have_std || a->non_resident || + len < offsetof(struct attr_record, + data.resident.reserved) + 1) + break; + value_len = le32_to_cpu(a->data.resident.value_length); + value_ofs = le16_to_cpu(a->data.resident.value_offset); + if (value_ofs > len || value_len > len - value_ofs) + break; + have_std = true; + } + a = (struct attr_record *)((u8 *)a + len); + } +out: + kfree(m); + return available; +free_m: + kfree(m); + return false; +} + +static s64 mft_reserve_end(const u8 *buf, s64 buf_start, s64 buf_end, + s64 start, s64 pass_end, s64 initialized_mft_records) +{ + s64 end = start + 1; + s64 limit = min_t(s64, start + MFT_RECORD_RESERVE, pass_end); + + if (limit > initialized_mft_records) + limit = initialized_mft_records; + if (limit > buf_end) + limit = buf_end; + while (end < limit && + !(buf[(end - buf_start) >> 3] & + (1 << ((end - buf_start) & 7)))) + end++; + return end; +} + +/* + * mft_bitmap_alloc_free_rec - find and allocate a free MFT record * @vol: volume on which to search for a free mft record * @base_ni: open base inode if allocating an extent mft record or NULL + * @max_mft_no: first record which must not be allocated, or -1 + * @new_reserve_end: if not NULL, end of a free run starting after the result * * Search for a free mft record in the mft bitmap attribute on the ntfs volume * @vol. @@ -862,10 +973,12 @@ static const char *es = " Leaving inconsistent metadata. Unmount and run chkds * * Locking: Caller must hold vol->mftbmp_lock for writing. */ -static s64 ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(struct ntfs_volume *vol, - struct ntfs_inode *base_ni) +static s64 mft_bitmap_alloc_free_rec(struct ntfs_volume *vol, + struct ntfs_inode *base_ni, + s64 max_mft_no, s64 *new_reserve_end) { s64 pass_end, ll, data_pos, pass_start, ofs, bit; + s64 initialized_mft_records; unsigned long flags; struct address_space *mftbmp_mapping; u8 *buf = NULL, *byte; @@ -882,30 +995,36 @@ static s64 ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(struct ntfs_volume *vo read_lock_irqsave(&NTFS_I(vol->mft_ino)->size_lock, flags); pass_end = NTFS_I(vol->mft_ino)->allocated_size >> vol->mft_record_size_bits; + initialized_mft_records = NTFS_I(vol->mft_ino)->initialized_size >> + vol->mft_record_size_bits; read_unlock_irqrestore(&NTFS_I(vol->mft_ino)->size_lock, flags); read_lock_irqsave(&NTFS_I(vol->mftbmp_ino)->size_lock, flags); ll = NTFS_I(vol->mftbmp_ino)->initialized_size << 3; read_unlock_irqrestore(&NTFS_I(vol->mftbmp_ino)->size_lock, flags); if (pass_end > ll) pass_end = ll; - pass = 1; - if (!base_ni) - data_pos = vol->mft_data_pos; - else - data_pos = base_ni->mft_no + 1; - if (data_pos < RESERVED_MFT_RECORDS) - data_pos = RESERVED_MFT_RECORDS; - if (data_pos >= pass_end) { - data_pos = RESERVED_MFT_RECORDS; + if (max_mft_no >= 0 && pass_end > max_mft_no) + pass_end = max_mft_no; + if (base_ni && base_ni->mft_no == FILE_MFT) { + data_pos = FILE_first_user; pass = 2; - /* This happens on a freshly formatted volume. */ if (data_pos >= pass_end) return -ENOSPC; - } - - if (base_ni && base_ni->mft_no == FILE_MFT) { - data_pos = 0; - pass = 2; + } else { + pass = 1; + if (!base_ni) + data_pos = vol->mft_data_pos; + else + data_pos = base_ni->mft_no + 1; + if (data_pos < FIRST_NORMAL_MFT_RECORD) + data_pos = FIRST_NORMAL_MFT_RECORD; + if (data_pos >= pass_end) { + data_pos = FIRST_NORMAL_MFT_RECORD; + pass = 2; + /* This happens on a freshly formatted volume. */ + if (data_pos >= pass_end) + return -ENOSPC; + } } pass_start = data_pos; @@ -940,38 +1059,28 @@ static s64 ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(struct ntfs_volume *vo size, data_pos, bit); for (; bit < size && data_pos + bit < pass_end; bit &= ~7ull, bit += 8) { - /* - * If we're extending $MFT and running out of the first - * mft record (base record) then give up searching since - * no guarantee that the found record will be accessible. - */ - if (base_ni && base_ni->mft_no == FILE_MFT && bit > 400) { - folio_unlock(folio); - kunmap_local(buf); - folio_put(folio); - return -ENOSPC; - } - byte = buf + (bit >> 3); if (*byte == 0xff) continue; - b = ffz((unsigned long)*byte); - if (b < 8 && b >= (bit & 7)) { + b = bit & 7; + for (; b < 8; b++) { + if (*byte & (1 << b)) + continue; ll = data_pos + (bit & ~7ull) + b; + if (ll >= pass_end) + break; + /* Keep the dynamic tail reserve for $MFT metadata. */ + if ((!base_ni || base_ni->mft_no != FILE_MFT) && + ll >= vol->mft_record_reserve_pos && + ll < vol->mft_record_reserve_end) + continue; if (unlikely(ll >= (1ll << 32))) { folio_unlock(folio); kunmap_local(buf); folio_put(folio); return -ENOSPC; } - *byte |= 1 << b; - folio_mark_dirty(folio); - folio_unlock(folio); - kunmap_local(buf); - folio_put(folio); - ntfs_debug("Done. (Found and allocated mft record 0x%llx.)", - ll); - return ll; + goto found; } } ntfs_debug("After inner for loop: size 0x%x, data_pos 0x%llx, bit 0x%llx", @@ -994,7 +1103,8 @@ static s64 ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(struct ntfs_volume *vo * part of the zone which we omitted earlier. */ pass_end = pass_start; - data_pos = pass_start = RESERVED_MFT_RECORDS; + data_pos = FIRST_NORMAL_MFT_RECORD; + pass_start = FIRST_NORMAL_MFT_RECORD; ntfs_debug("pass %i, pass_start 0x%llx, pass_end 0x%llx.", pass, pass_start, pass_end); if (data_pos >= pass_end) @@ -1004,9 +1114,22 @@ static s64 ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(struct ntfs_volume *vo /* No free mft records in currently initialized mft bitmap. */ ntfs_debug("Done. (No free mft records left in currently initialized mft bitmap.)"); return -ENOSPC; +found: + if (new_reserve_end) + *new_reserve_end = mft_reserve_end(buf, data_pos, + data_pos + size, ll, pass_end, + initialized_mft_records); + *byte |= 1 << b; + folio_mark_dirty(folio); + folio_unlock(folio); + kunmap_local(buf); + folio_put(folio); + ntfs_debug("Done. (Found and allocated mft record 0x%llx.)", ll); + return ll; } -static int ntfs_mft_attr_extend(struct ntfs_inode *ni) +static int ntfs_mft_attr_extend(struct ntfs_inode *ni, + struct ntfs_inode *locked_ni) { int ret = 0; struct ntfs_inode *base_ni; @@ -1027,7 +1150,7 @@ static int ntfs_mft_attr_extend(struct ntfs_inode *ni) } } - ret = ntfs_attr_update_mapping_pairs(ni, 0); + ret = ntfs_attr_update_mapping_pairs_locked(ni, 0, locked_ni); if (ret) pr_err("MP update failed\n"); @@ -1215,7 +1338,7 @@ static int ntfs_mft_bitmap_extend_allocation_nolock(struct ntfs_volume *vol) ret = ntfs_attr_record_resize(ctx->mrec, a, mp_size + le16_to_cpu(a->data.non_resident.mapping_pairs_offset)); if (unlikely(ret)) { - ret = ntfs_mft_attr_extend(mftbmp_ni); + ret = ntfs_mft_attr_extend(mftbmp_ni, mftbmp_ni); if (!ret) goto extended_ok; if (ret != -EAGAIN) @@ -1326,7 +1449,9 @@ undo_alloc: NVolSetErrors(vol); } mark_mft_record_dirty(ctx->ntfs_ino); - } else if (status.mp_extended && ntfs_attr_update_mapping_pairs(mftbmp_ni, 0)) { + } else if (status.mp_extended && + ntfs_attr_update_mapping_pairs_locked(mftbmp_ni, 0, + mftbmp_ni)) { ntfs_error(vol->sb, "Failed to restore mapping pairs.%s", es); NVolSetErrors(vol); } @@ -1414,7 +1539,6 @@ static int ntfs_mft_bitmap_extend_initialized_nolock(struct ntfs_volume *vol) ret = ntfs_attr_set(mftbmp_ni, old_initialized_size, 8, 0); if (likely(!ret)) { ntfs_debug("Done. (Wrote eight initialized bytes to mft bitmap."); - ntfs_inc_free_mft_records(vol, 8 * 8); return 0; } ntfs_error(vol->sb, "Failed to write to mft bitmap."); @@ -1471,8 +1595,9 @@ err_out: * @vol: volume on which to extend the mft data attribute * * Extend the mft data attribute on the ntfs volume @vol by 16 mft records - * worth of clusters or if not enough space for this by one mft record worth - * of clusters. + * worth of clusters or if not enough space for this by two mft records worth + * of clusters. Keeping at least two new records breaks the recursion between + * extending $MFT and allocating a record for a new $MFT attribute extent. * * Note: Only changes allocated_size, i.e. does not touch initialized_size or * data_size. @@ -1526,10 +1651,8 @@ static int ntfs_mft_data_extend_allocation_nolock(struct ntfs_volume *vol) } lcn = rl->lcn + rl->length; ntfs_debug("Last lcn of mft data attribute is 0x%llx.", lcn); - /* Minimum allocation is one mft record worth of clusters. */ - min_nr = NTFS_B_TO_CLU(vol, vol->mft_record_size); - if (!min_nr) - min_nr = 1; + /* Keep room for the allocating record and at least one MFT reserve. */ + min_nr = DIV_ROUND_UP_ULL((u64)vol->mft_record_size * 2, vol->cluster_size); /* Want to allocate 16 mft records worth of clusters. */ nr = vol->mft_record_size << 4 >> vol->cluster_size_bits; if (!nr) @@ -1653,7 +1776,7 @@ static int ntfs_mft_data_extend_allocation_nolock(struct ntfs_volume *vol) ret = ntfs_attr_record_resize(ctx->mrec, a, mp_size + le16_to_cpu(a->data.non_resident.mapping_pairs_offset)); if (unlikely(ret)) { - ret = ntfs_mft_attr_extend(mft_ni); + ret = ntfs_mft_attr_extend(mft_ni, NULL); if (!ret) goto extended_ok; if (ret != -EAGAIN) @@ -1928,6 +2051,7 @@ static int ntfs_mft_record_format(const struct ntfs_volume *vol, const s64 mft_n * @ni: [OUT] on success, set to the allocated ntfs inode * @base_ni: [IN] open base inode if allocating an extent mft record or NULL * @ni_mrec: [OUT] on successful return this is the mapped mft record + * @mft_data_vcn: [IN] lowest VCN of a new $MFT/$DATA extent, or -1 * * Allocate an mft record in $MFT/$DATA of an open ntfs volume @vol. * @@ -1955,30 +2079,23 @@ static int ntfs_mft_record_format(const struct ntfs_volume *vol, const s64 mft_n * optimize this we start scanning at the place specified by @base_ni or if * @base_ni is NULL we start where we last stopped and we perform wrap around * when we reach the end. Note, we do not try to allocate mft records below - * number 64 because numbers 0 to 15 are the defined system files anyway and 16 - * to 64 are special in that they are used for storing extension mft records - * for the $DATA attribute of $MFT. This is required to avoid the possibility - * of creating a runlist with a circular dependency which once written to disk - * can never be read in again. Windows will only use records 16 to 24 for - * normal files if the volume is completely out of space. We never use them - * which means that when the volume is really out of space we cannot create any - * more files while Windows can still create up to 8 small files. We can start - * doing this at some later time, it does not matter much for now. + * number 24 because numbers 0 to 15 are the defined system files and records + * 16 to 23 are kept for metadata compatibility. Records reserved dynamically + * at the initialized MFT tail are skipped by normal allocation and consumed by + * $MFT metadata extent allocation. * * When scanning the mft bitmap, we only search up to the last allocated mft - * record. If there are no free records left in the range 64 to number of + * record. If there are no free records left in the range 24 to number of * allocated mft records, then we extend the $MFT/$DATA attribute in order to * create free mft records. We extend the allocated size of $MFT/$DATA by 16 * records at a time or one cluster, if cluster size is above 16kiB. If there - * is not sufficient space to do this, we try to extend by a single mft record - * or one cluster, if cluster size is above the mft record size. + * is not sufficient space to do this, we try to extend by two mft records or + * one cluster, if a cluster already contains at least two mft records. * - * No matter how many mft records we allocate, we initialize only the first - * allocated mft record, incrementing mft data size and initialized size - * accordingly, open an struct ntfs_inode for it and return it to the caller, unless - * there are less than 64 mft records, in which case we allocate and initialize - * mft records until we reach record 64 which we consider as the first free mft - * record for use by normal files. + * When extending the initialized MFT tail, we also initialize up to four + * additional records and reserve them in memory for future $MFT metadata + * extents. If there are less than 24 mft records, records are initialized + * until record 24, which is the first record used for normal files. * * If during any stage we overflow the initialized data in the mft bitmap, we * extend the initialized size (and data size) by 8 bytes, allocating another @@ -2014,9 +2131,13 @@ static int ntfs_mft_record_format(const struct ntfs_volume *vol, const s64 mft_n */ int ntfs_mft_record_alloc(struct ntfs_volume *vol, const int mode, struct ntfs_inode **ni, struct ntfs_inode *base_ni, - struct mft_record **ni_mrec) + struct mft_record **ni_mrec, const s64 mft_data_vcn) { s64 ll, bit, old_data_initialized, old_data_size; + s64 nr_new_mft_records = 0; + s64 max_mft_no = -1, reserve_start = -1, reserve_end = -1; + s64 candidate_reserve_end = -1; + s64 *reserve_endp; unsigned long flags; struct folio *folio; struct ntfs_inode *mft_ni, *mftbmp_ni; @@ -2027,7 +2148,9 @@ int ntfs_mft_record_alloc(struct ntfs_volume *vol, const int mode, unsigned int ofs; int err; __le16 seq_no, usn; - bool record_formatted = false; + bool record_formatted = false, from_reserve = false, tail_alloc = false; + bool reserve_created = false; + bool forced_reserved_record = false; unsigned int memalloc_flags; if (base_ni && *ni) @@ -2036,6 +2159,21 @@ int ntfs_mft_record_alloc(struct ntfs_volume *vol, const int mode, /* @mode and @base_ni are mutually exclusive. */ if (mode && base_ni) return -EINVAL; + if (mft_data_vcn >= 0 && + (!base_ni || base_ni->mft_no != FILE_MFT)) + return -EINVAL; + if (mft_data_vcn >= 0) { + u64 vbo; + + if ((u64)mft_data_vcn > (U64_MAX >> vol->cluster_size_bits)) + return -EOVERFLOW; + vbo = (u64)mft_data_vcn << vol->cluster_size_bits; + /* + * The whole extent record must be reachable without this + * extent, including when an MFT record spans multiple clusters. + */ + max_mft_no = vbo >> vol->mft_record_size_bits; + } if (base_ni) ntfs_debug("Entering (allocating an extent mft record for base mft record 0x%llx).", @@ -2050,10 +2188,39 @@ int ntfs_mft_record_alloc(struct ntfs_volume *vol, const int mode, mutex_lock(&mft_ni->mrec_lock); mftbmp_ni = NTFS_I(vol->mftbmp_ino); search_free_rec: + from_reserve = false; + reserve_created = false; + candidate_reserve_end = -1; if (!base_ni || base_ni->mft_no != FILE_MFT) down_write(&vol->mftbmp_lock); - bit = ntfs_mft_bitmap_find_and_alloc_free_rec_nolock(vol, base_ni); + if (base_ni && base_ni->mft_no == FILE_MFT && + vol->mft_record_reserve_pos < vol->mft_record_reserve_end && + (max_mft_no < 0 || vol->mft_record_reserve_pos < max_mft_no)) { + bit = vol->mft_record_reserve_pos; + err = ntfs_bitmap_set_bit(vol->mftbmp_ino, bit); + if (unlikely(err)) { + ntfs_error(vol->sb, + "Failed to allocate reserved MFT record 0x%llx.", + bit); + goto err_out; + } + vol->mft_record_reserve_pos++; + from_reserve = true; + ntfs_debug("Allocated MFT metadata record 0x%llx from tail reserve.", + bit); + goto have_alloc_rec; + } + reserve_endp = vol->mft_record_reserve_pos >= + vol->mft_record_reserve_end ? &candidate_reserve_end : NULL; + bit = mft_bitmap_alloc_free_rec(vol, base_ni, max_mft_no, reserve_endp); if (bit >= 0) { + if (candidate_reserve_end > bit + 1) { + vol->mft_record_reserve_pos = bit + 1; + vol->mft_record_reserve_end = candidate_reserve_end; + reserve_created = true; + ntfs_debug("Reserved free MFT records [0x%llx, 0x%llx) for metadata.", + bit + 1, candidate_reserve_end); + } ntfs_debug("Found and allocated free record (#1), bit 0x%llx.", (long long)bit); goto have_alloc_rec; @@ -2068,6 +2235,24 @@ search_free_rec: } if (base_ni && base_ni->mft_no == FILE_MFT) { + static const u8 bootstrap_records[] = { + FILE_reserved15, FILE_reserved12, FILE_reserved13, + FILE_reserved14, + }; + int i; + + for (i = 0; i < ARRAY_SIZE(bootstrap_records); i++) { + if (max_mft_no >= 0 && bootstrap_records[i] >= max_mft_no) + continue; + if (!mft_reserved_is_free(vol, mft_ni, + bootstrap_records[i])) + continue; + bit = bootstrap_records[i]; + forced_reserved_record = true; + ntfs_debug("Using reserved MFT record %lld to bootstrap metadata extension.", + bit); + goto have_alloc_rec; + } memalloc_nofs_restore(memalloc_flags); return bit; } @@ -2087,10 +2272,10 @@ search_free_rec: old_data_initialized = mftbmp_ni->initialized_size; read_unlock_irqrestore(&mftbmp_ni->size_lock, flags); if (old_data_initialized << 3 > ll && - old_data_initialized > RESERVED_MFT_RECORDS / 8) { + old_data_initialized << 3 > FIRST_NORMAL_MFT_RECORD) { bit = ll; - if (bit < RESERVED_MFT_RECORDS) - bit = RESERVED_MFT_RECORDS; + if (bit < FIRST_NORMAL_MFT_RECORD) + bit = FIRST_NORMAL_MFT_RECORD; if (unlikely(bit >= (1ll << 32))) goto max_err_out; ntfs_debug("Found free record (#2), bit 0x%llx.", @@ -2176,6 +2361,11 @@ have_alloc_rec: read_lock_irqsave(&mft_ni->size_lock, flags); old_data_initialized = mft_ni->initialized_size; read_unlock_irqrestore(&mft_ni->size_lock, flags); + tail_alloc = (!base_ni || base_ni->mft_no != FILE_MFT) && + bit >= (old_data_initialized >> vol->mft_record_size_bits) && + vol->mft_record_reserve_pos >= vol->mft_record_reserve_end; + if (tail_alloc) + ll = (bit + 2) << vol->mft_record_size_bits; if (ll <= old_data_initialized) { ntfs_debug("Allocated mft record already initialized."); goto mft_rec_already_initialized; @@ -2208,6 +2398,29 @@ have_alloc_rec: mft_ni->initialized_size); } read_unlock_irqrestore(&mft_ni->size_lock, flags); + if (tail_alloc) { + s64 bitmap_records; + + read_lock_irqsave(&mft_ni->size_lock, flags); + reserve_end = mft_ni->allocated_size >> + vol->mft_record_size_bits; + read_unlock_irqrestore(&mft_ni->size_lock, flags); + read_lock_irqsave(&mftbmp_ni->size_lock, flags); + bitmap_records = mftbmp_ni->initialized_size << 3; + read_unlock_irqrestore(&mftbmp_ni->size_lock, flags); + if (reserve_end > bitmap_records) + reserve_end = bitmap_records; + if (reserve_end > bit + 1 + MFT_RECORD_RESERVE) + reserve_end = bit + 1 + MFT_RECORD_RESERVE; + reserve_start = bit + 1; + if (reserve_end > reserve_start) { + ll = reserve_end << vol->mft_record_size_bits; + } else { + reserve_start = -1; + reserve_end = -1; + ll = (bit + 1) << vol->mft_record_size_bits; + } + } } else if (ll > mft_ni->allocated_size) { err = -ENOSPC; goto undo_mftbmp_alloc_nolock; @@ -2276,14 +2489,25 @@ have_alloc_rec: mark_mft_record_dirty(ctx->ntfs_ino); ntfs_attr_put_search_ctx(ctx); unmap_mft_record(mft_ni); + if (reserve_start >= 0 && reserve_end > reserve_start) { + vol->mft_record_reserve_pos = reserve_start; + vol->mft_record_reserve_end = reserve_end; + ntfs_debug("Reserved MFT records [0x%llx, 0x%llx) for metadata.", + reserve_start, reserve_end); + } read_lock_irqsave(&mft_ni->size_lock, flags); ntfs_debug("Status of mft data after mft record initialization: allocated_size 0x%llx, data_size 0x%llx, initialized_size 0x%llx.", mft_ni->allocated_size, i_size_read(vol->mft_ino), mft_ni->initialized_size); WARN_ON(i_size_read(vol->mft_ino) > mft_ni->allocated_size); WARN_ON(mft_ni->initialized_size > i_size_read(vol->mft_ino)); + nr_new_mft_records = (i_size_read(vol->mft_ino) - old_data_size) >> + vol->mft_record_size_bits; read_unlock_irqrestore(&mft_ni->size_lock, flags); mft_rec_already_initialized: + /* Account for newly visible MFT records before dropping the lock. */ + if (nr_new_mft_records > 0) + ntfs_inc_free_mft_records(vol, nr_new_mft_records); /* * We can finally drop the mft bitmap lock as the mft data attribute * has been fully updated. The only disparity left is that the @@ -2315,8 +2539,8 @@ mft_rec_already_initialized: /* If we just formatted the mft record no need to do it again. */ if (!record_formatted) { /* Sanity check that the mft record is really not in use. */ - if (ntfs_is_file_record(m->magic) && - (m->flags & MFT_RECORD_IN_USE)) { + if (!forced_reserved_record && ntfs_is_file_record(m->magic) && + (m->flags & MFT_RECORD_IN_USE)) { ntfs_warning(vol->sb, "Mft record 0x%llx was marked free in mft bitmap but is marked used itself. Unmount and run chkdsk.", bit); @@ -2389,9 +2613,13 @@ mft_rec_already_initialized: ntfs_error(vol->sb, "Failed to map allocated extent mft record 0x%llx.", bit); err = PTR_ERR(m_tmp); - /* Set the mft record itself not in use. */ - m->flags &= cpu_to_le16( - ~le16_to_cpu(MFT_RECORD_IN_USE)); + if (forced_reserved_record) { + m->base_mft_record = 0; + m->flags |= MFT_RECORD_IN_USE; + } else { + /* Set the mft record itself not in use. */ + m->flags &= cpu_to_le16(~le16_to_cpu(MFT_RECORD_IN_USE)); + } /* Make sure the mft record is written out to disk. */ ntfs_mft_mark_dirty(folio); folio_unlock(folio); @@ -2463,7 +2691,8 @@ mft_rec_already_initialized: (*ni)->mft_no = bit; if (ni_mrec) *ni_mrec = (*ni)->mrec; - ntfs_dec_free_mft_records(vol, 1); + if (!forced_reserved_record) + ntfs_dec_free_mft_records(vol, 1); return 0; undo_data_init: write_lock_irqsave(&mft_ni->size_lock, flags); @@ -2475,10 +2704,13 @@ undo_mftbmp_alloc: if (!base_ni || base_ni->mft_no != FILE_MFT) down_write(&vol->mftbmp_lock); undo_mftbmp_alloc_nolock: - if (ntfs_bitmap_clear_bit(vol->mftbmp_ino, bit)) { + if (!forced_reserved_record && ntfs_bitmap_clear_bit(vol->mftbmp_ino, bit)) { ntfs_error(vol->sb, "Failed to clear bit in mft bitmap.%s", es); NVolSetErrors(vol); } + if ((from_reserve || reserve_created) && + vol->mft_record_reserve_pos == bit + 1) + vol->mft_record_reserve_pos = bit; if (!base_ni || base_ni->mft_no != FILE_MFT) up_write(&vol->mftbmp_lock); err_out: @@ -2514,9 +2746,11 @@ int ntfs_mft_record_free(struct ntfs_volume *vol, struct ntfs_inode *ni) int err; u16 seq_no; __le16 old_seq_no; + __le64 old_base_mft_record; struct mft_record *ni_mrec; unsigned int memalloc_flags; struct ntfs_inode *base_ni; + bool keep_reserved; if (!vol || !ni) return -EINVAL; @@ -2529,9 +2763,23 @@ int ntfs_mft_record_free(struct ntfs_volume *vol, struct ntfs_inode *ni) /* Cache the mft reference for later. */ mft_no = ni->mft_no; - - /* Mark the mft record as not in use. */ - ni_mrec->flags &= ~MFT_RECORD_IN_USE; + if (likely(ni->nr_extents >= 0)) + base_ni = ni; + else + base_ni = ni->ext.base_ntfs_ino; + keep_reserved = mft_no >= FILE_reserved12 && + mft_no <= FILE_reserved15 && + base_ni->mft_no == FILE_MFT; + + old_base_mft_record = ni_mrec->base_mft_record; + if (keep_reserved) { + /* Restore the special, unnamed form used by reserved records. */ + ni_mrec->base_mft_record = 0; + ni_mrec->flags |= MFT_RECORD_IN_USE; + } else { + /* Mark the mft record as not in use. */ + ni_mrec->flags &= ~MFT_RECORD_IN_USE; + } /* Increment the sequence number, skipping zero, if it is not zero. */ old_seq_no = ni_mrec->sequence_number; @@ -2560,24 +2808,28 @@ int ntfs_mft_record_free(struct ntfs_volume *vol, struct ntfs_inode *ni) if (err) goto sync_rollback; - if (likely(ni->nr_extents >= 0)) - base_ni = ni; - else - base_ni = ni->ext.base_ntfs_ino; + if (keep_reserved) { + unmap_mft_record(ni); + return 0; + } /* Clear the bit in the $MFT/$BITMAP corresponding to this record. */ memalloc_flags = memalloc_nofs_save(); if (base_ni->mft_no != FILE_MFT) down_write(&vol->mftbmp_lock); err = ntfs_bitmap_clear_bit(vol->mftbmp_ino, mft_no); + if (!err) + ntfs_inc_free_mft_records(vol, 1); + if (!err && base_ni->mft_no == FILE_MFT && + mft_no + 1 == vol->mft_record_reserve_pos && + mft_no < vol->mft_record_reserve_end) + vol->mft_record_reserve_pos = mft_no; if (base_ni->mft_no != FILE_MFT) up_write(&vol->mftbmp_lock); memalloc_nofs_restore(memalloc_flags); if (err) goto bitmap_rollback; - unmap_mft_record(ni); - ntfs_inc_free_mft_records(vol, 1); return 0; /* Rollback what we did... */ @@ -2595,6 +2847,7 @@ sync_rollback: "Eeek! Rollback failed in %s. Leaving inconsistent metadata!\n", __func__); ni_mrec->flags |= MFT_RECORD_IN_USE; ni_mrec->sequence_number = old_seq_no; + ni_mrec->base_mft_record = old_base_mft_record; NInoSetDirty(ni); write_mft_record(ni, ni_mrec, 0); unmap_mft_record(ni); diff --git a/fs/ntfs/mft.h b/fs/ntfs/mft.h index 75a51a98d0f6..ed5c1d595c0d 100644 --- a/fs/ntfs/mft.h +++ b/fs/ntfs/mft.h @@ -78,7 +78,7 @@ static inline int write_mft_record(struct ntfs_inode *ni, struct mft_record *m, int ntfs_mft_record_alloc(struct ntfs_volume *vol, const int mode, struct ntfs_inode **ni, struct ntfs_inode *base_ni, - struct mft_record **ni_mrec); + struct mft_record **ni_mrec, const s64 mft_data_vcn); int ntfs_mft_record_free(struct ntfs_volume *vol, struct ntfs_inode *ni); int ntfs_mft_records_write(const struct ntfs_volume *vol, const u64 mref, const s64 count, struct mft_record *b); diff --git a/fs/ntfs/namei.c b/fs/ntfs/namei.c index 7091b2496fac..fdf52fac4329 100644 --- a/fs/ntfs/namei.c +++ b/fs/ntfs/namei.c @@ -480,7 +480,7 @@ static struct ntfs_inode *__ntfs_create(struct mnt_idmap *idmap, struct inode *d mark_inode_dirty(dir); err = ntfs_mft_record_alloc(dir_ni->vol, mode, &ni, NULL, - &ni_mrec); + &ni_mrec, -1); if (err) { iput(vi); return ERR_PTR(err); diff --git a/fs/ntfs/super.c b/fs/ntfs/super.c index 5aad2d2a36bb..4066bacabe37 100644 --- a/fs/ntfs/super.c +++ b/fs/ntfs/super.c @@ -2064,8 +2064,7 @@ static unsigned long __get_nr_free_mft_records(struct ntfs_volume *vol, /* If errors occurred we may well have gone below zero, fix this. */ if (nr_free < 0) nr_free = 0; - else - atomic64_set(&vol->free_mft_records, nr_free); + atomic64_set(&vol->free_mft_records, nr_free); ntfs_debug("Exiting."); return nr_free; @@ -2131,7 +2130,14 @@ static int ntfs_statfs(struct dentry *dentry, struct kstatfs *sfs) read_unlock_irqrestore(&mft_ni->size_lock, flags); /* Free inodes in fs (based on current total count). */ - sfs->f_ffree = atomic64_read(&vol->free_mft_records); + size = atomic64_read(&vol->free_mft_records); + if (unlikely(size < 0 || size > (s64)sfs->f_files)) + ntfs_warning(vol->sb, "Invalid free MFT record count %lld.", size); + if (size < 0) + size = 0; + else if (size > (s64)sfs->f_files) + size = sfs->f_files; + sfs->f_ffree = size; /* * File system id. This is extremely *nix flavour dependent and even diff --git a/fs/ntfs/volume.h b/fs/ntfs/volume.h index 65fd3908af26..bc85a9592245 100644 --- a/fs/ntfs/volume.h +++ b/fs/ntfs/volume.h @@ -55,6 +55,10 @@ * @attrdef_size: Size of the attribute definition table in bytes. * @attrdef: Table of attribute definitions. Obtained from FILE_AttrDef. * @mft_data_pos: Mft record number at which to allocate the next mft record. + * @mft_record_reserve_pos: First record in the in-memory MFT metadata reserve + * (protected by mftbmp_lock). + * @mft_record_reserve_end: First record beyond the MFT metadata reserve + * (protected by mftbmp_lock). * @mft_zone_start: First cluster of the mft zone. * @mft_zone_end: First cluster beyond the mft zone. * @mft_zone_pos: Current position in the mft zone. @@ -119,6 +123,8 @@ struct ntfs_volume { s32 attrdef_size; struct attr_def *attrdef; s64 mft_data_pos; + s64 mft_record_reserve_pos; + s64 mft_record_reserve_end; s64 mft_zone_start; s64 mft_zone_end; s64 mft_zone_pos; @@ -252,17 +258,11 @@ static inline void ntfs_dec_free_clusters(struct ntfs_volume *vol, s64 nr) static inline void ntfs_inc_free_mft_records(struct ntfs_volume *vol, s64 nr) { - if (!NVolFreeClusterKnown(vol)) - return; - atomic64_add(nr, &vol->free_mft_records); } static inline void ntfs_dec_free_mft_records(struct ntfs_volume *vol, s64 nr) { - if (!NVolFreeClusterKnown(vol)) - return; - atomic64_sub(nr, &vol->free_mft_records); } diff --git a/fs/smb/client/cifssmb.c b/fs/smb/client/cifssmb.c index f9aff0712794..6dddbd84b93b 100644 --- a/fs/smb/client/cifssmb.c +++ b/fs/smb/client/cifssmb.c @@ -3080,7 +3080,7 @@ int cifs_query_reparse_point(const unsigned int xid, end = 2 + get_bcc(&io_rsp->hdr) + (__u8 *)&io_rsp->ByteCount; start = (__u8 *)&io_rsp->hdr.Protocol + data_offset; - if (start >= end) { + if (start >= end || (size_t)(end - start) < sizeof(*buf)) { rc = smb_EIO2(smb_eio_trace_qreparse_data_area, (unsigned long)start - (unsigned long)io_rsp, (unsigned long)end - (unsigned long)io_rsp); diff --git a/fs/smb/client/connect.c b/fs/smb/client/connect.c index b6e98eb31673..28e1ddeb6182 100644 --- a/fs/smb/client/connect.c +++ b/fs/smb/client/connect.c @@ -174,6 +174,8 @@ cifs_signal_cifsd_for_reconnect(struct TCP_Server_Info *server, nserver = ses->chans[i].server; if (!nserver) continue; + if (!list_empty(&nserver->rlist)) + continue; nserver->srv_count++; list_add(&nserver->rlist, &reco); } @@ -182,11 +184,15 @@ cifs_signal_cifsd_for_reconnect(struct TCP_Server_Info *server, } } + spin_lock(&cifs_tcp_ses_lock); list_for_each_entry_safe(server, nserver, &reco, rlist) { list_del_init(&server->rlist); set_need_reco(server); + spin_unlock(&cifs_tcp_ses_lock); cifs_put_tcp_session(server, 0); + spin_lock(&cifs_tcp_ses_lock); } + spin_unlock(&cifs_tcp_ses_lock); } /* @@ -1067,6 +1073,7 @@ clean_demultiplex_info(struct TCP_Server_Info *server) spin_unlock(&server->srv_lock); cancel_delayed_work_sync(&server->echo); + cancel_delayed_work_sync(&server->reconnect); spin_lock(&server->srv_lock); server->tcpStatus = CifsExiting; @@ -1823,6 +1830,7 @@ cifs_get_tcp_session(struct smb3_fs_context *ctx, spin_lock_init(&tcp_ses->mid_counter_lock); INIT_LIST_HEAD(&tcp_ses->tcp_ses_list); INIT_LIST_HEAD(&tcp_ses->smb_ses_list); + INIT_LIST_HEAD(&tcp_ses->rlist); INIT_DELAYED_WORK(&tcp_ses->echo, cifs_echo_request); INIT_DELAYED_WORK(&tcp_ses->reconnect, smb2_reconnect_server); mutex_init(&tcp_ses->reconnect_mutex); @@ -1926,6 +1934,7 @@ out_err: kfree(tcp_ses->leaf_fullpath); if (tcp_ses->ssocket) sock_release(tcp_ses->ssocket); + smbd_destroy(tcp_ses); kfree(tcp_ses); } return ERR_PTR(rc); diff --git a/fs/smb/client/file.c b/fs/smb/client/file.c index 1aa4844f8b8a..0d428517f454 100644 --- a/fs/smb/client/file.c +++ b/fs/smb/client/file.c @@ -3354,8 +3354,8 @@ void cifs_oplock_break(struct work_struct *work) wait_on_bit(&cinode->flags, CIFS_INODE_PENDING_WRITERS, TASK_UNINTERRUPTIBLE); - tlink = cifs_sb_tlink(cifs_sb); - if (IS_ERR(tlink)) { + tlink = cifs_get_tlink(cfile->tlink); + if (IS_ERR_OR_NULL(tlink)) { /* drop the reference taken when the break was queued */ _cifsFileInfo_put(cfile, false /* do not wait for ourself */, false); goto out; diff --git a/fs/smb/client/misc.c b/fs/smb/client/misc.c index 945194fe7a97..05168284f205 100644 --- a/fs/smb/client/misc.c +++ b/fs/smb/client/misc.c @@ -788,7 +788,11 @@ parse_dfs_referrals(struct get_dfs_referral_rsp *rsp, u32 rsp_size, node->ref_flag = le16_to_cpu(ref->ReferralEntryFlags); /* copy DfsPath */ - if (le16_to_cpu(ref->DfsPathOffset) > data_end - (char *)ref) { + if (le16_to_cpu(ref->DfsPathOffset) < sizeof(*ref) || + le16_to_cpu(ref->DfsPathOffset) > data_end - (char *)ref) { + cifs_dbg(VFS, "%s: DfsPathOffset %u out of range [%zu, %td]\n", + __func__, le16_to_cpu(ref->DfsPathOffset), + sizeof(*ref), data_end - (char *)ref); rc = -EINVAL; goto parse_DFS_referrals_exit; } @@ -802,7 +806,11 @@ parse_dfs_referrals(struct get_dfs_referral_rsp *rsp, u32 rsp_size, } /* copy link target UNC */ - if (le16_to_cpu(ref->NetworkAddressOffset) > data_end - (char *)ref) { + if (le16_to_cpu(ref->NetworkAddressOffset) < sizeof(*ref) || + le16_to_cpu(ref->NetworkAddressOffset) > data_end - (char *)ref) { + cifs_dbg(VFS, "%s: NetworkAddressOffset %u out of range [%zu, %td]\n", + __func__, le16_to_cpu(ref->NetworkAddressOffset), + sizeof(*ref), data_end - (char *)ref); rc = -EINVAL; goto parse_DFS_referrals_exit; } diff --git a/fs/smb/client/reparse.c b/fs/smb/client/reparse.c index 8a1b9e8be5ba..3a27773186ae 100644 --- a/fs/smb/client/reparse.c +++ b/fs/smb/client/reparse.c @@ -3,6 +3,7 @@ * Copyright (c) 2024 Paulo Alcantara <pc@manguebit.com> */ +#include <linux/ctype.h> #include <linux/fs.h> #include <linux/stat.h> #include <linux/slab.h> @@ -159,15 +160,24 @@ static int create_native_symlink(const unsigned int xid, struct inode *inode, convert_delimiter(sym, sep); /* - * For absolute NT symlinks it is required to pass also leading - * backslash and to not mangle NT object prefix "\\??\\" and not to - * mangle colon in drive letter. But cifs_convert_path_to_utf16() - * removes leading backslash and replaces '?' and ':'. So temporary - * mask these characters in NT object prefix by '_' and then change - * them back. + * Absolute NT symlinks must retain the leading backslash, "\\??\\" + * prefix and drive-letter colon. cifs_convert_path_to_utf16() strips + * the leading backslash and maps '?' and ':', so temporarily mask + * these characters with '_' and restore them after conversion. + * + * When symlinkroot is unset, sym comes directly from the caller. + * Validate the complete "\\??\\X:" prefix before using fixed offsets + * or subtracting the NT prefix length below. Require an ASCII drive + * letter so the prefix occupies six characters in UTF-16 too. */ - if (!(sbflags & CIFS_MOUNT_POSIX_PATHS) && symname[0] == '/') + if (!(sbflags & CIFS_MOUNT_POSIX_PATHS) && symname[0] == '/') { + if (!strstarts(sym, "\\??\\") || !isascii(sym[4]) || + !isalpha(sym[4]) || sym[5] != ':') { + rc = -EINVAL; + goto out; + } sym[0] = sym[1] = sym[2] = sym[5] = '_'; + } /* * On a POSIX paths mount the symlink target is stored verbatim, so @@ -1139,29 +1149,30 @@ static bool wsl_to_fattr(struct cifs_open_info_data *data, u32 tag, struct cifs_fattr *fattr) { unsigned int sbflags = cifs_sb_flags(cifs_sb); + kuid_t uid = cifs_sb->ctx->linux_uid; + kgid_t gid = cifs_sb->ctx->linux_gid; struct smb2_file_full_ea_info *ea; bool have_xattr_dev = false; + dev_t rdev = 0; + umode_t mode; u32 next = 0; - fattr->cf_uid = cifs_sb->ctx->linux_uid; - fattr->cf_gid = cifs_sb->ctx->linux_gid; - - fattr->cf_mode &= ~S_IFMT; + mode = fattr->cf_mode & ~S_IFMT; switch (tag) { case IO_REPARSE_TAG_LX_SYMLINK: - fattr->cf_mode |= S_IFLNK; + mode |= S_IFLNK; break; case IO_REPARSE_TAG_LX_FIFO: - fattr->cf_mode |= S_IFIFO; + mode |= S_IFIFO; break; case IO_REPARSE_TAG_AF_UNIX: - fattr->cf_mode |= S_IFSOCK; + mode |= S_IFSOCK; break; case IO_REPARSE_TAG_LX_CHR: - fattr->cf_mode |= S_IFCHR; + mode |= S_IFCHR; break; case IO_REPARSE_TAG_LX_BLK: - fattr->cf_mode |= S_IFBLK; + mode |= S_IFBLK; break; } @@ -1185,26 +1196,29 @@ static bool wsl_to_fattr(struct cifs_open_info_data *data, if (!strncmp(name, SMB2_WSL_XATTR_UID, nlen)) { if (!(sbflags & CIFS_MOUNT_OVERR_UID)) - fattr->cf_uid = wsl_make_kuid(cifs_sb, v); + uid = wsl_make_kuid(cifs_sb, v); } else if (!strncmp(name, SMB2_WSL_XATTR_GID, nlen)) { if (!(sbflags & CIFS_MOUNT_OVERR_GID)) - fattr->cf_gid = wsl_make_kgid(cifs_sb, v); + gid = wsl_make_kgid(cifs_sb, v); } else if (!strncmp(name, SMB2_WSL_XATTR_MODE, nlen)) { /* File type in reparse point tag and in xattr mode must match. */ - if (S_DT(fattr->cf_mode) != S_DT(le32_to_cpu(*(__le32 *)v))) + if (S_DT(mode) != S_DT(get_unaligned_le32(v))) return false; - fattr->cf_mode = (umode_t)le32_to_cpu(*(__le32 *)v); + mode = get_unaligned_le32(v); } else if (!strncmp(name, SMB2_WSL_XATTR_DEV, nlen)) { - fattr->cf_rdev = reparse_mkdev(v); + rdev = reparse_mkdev(v); have_xattr_dev = true; } } while (next); out: - /* Major and minor numbers for char and block devices are mandatory. */ if (!have_xattr_dev && (tag == IO_REPARSE_TAG_LX_CHR || tag == IO_REPARSE_TAG_LX_BLK)) return false; + fattr->cf_uid = uid; + fattr->cf_gid = gid; + fattr->cf_mode = mode; + fattr->cf_rdev = rdev; return true; } diff --git a/fs/smb/client/reparse.h b/fs/smb/client/reparse.h index 49efd85b1e94..05b2cecb4495 100644 --- a/fs/smb/client/reparse.h +++ b/fs/smb/client/reparse.h @@ -9,6 +9,7 @@ #include <linux/fs.h> #include <linux/stat.h> #include <linux/uidgid.h> +#include <linux/unaligned.h> #include "fs_context.h" #include "cifsglob.h" #include "../common/smbfsctl.h" @@ -23,7 +24,7 @@ static inline dev_t reparse_mkdev(void *ptr) { - u64 v = le64_to_cpu(*(__le64 *)ptr); + u64 v = get_unaligned_le64(ptr); return MKDEV(v & 0xffffffff, v >> 32); } @@ -31,7 +32,7 @@ static inline dev_t reparse_mkdev(void *ptr) static inline kuid_t wsl_make_kuid(struct cifs_sb_info *cifs_sb, void *ptr) { - u32 uid = le32_to_cpu(*(__le32 *)ptr); + u32 uid = get_unaligned_le32(ptr); if (cifs_sb_flags(cifs_sb) & CIFS_MOUNT_OVERR_UID) return cifs_sb->ctx->linux_uid; @@ -41,7 +42,7 @@ static inline kuid_t wsl_make_kuid(struct cifs_sb_info *cifs_sb, static inline kgid_t wsl_make_kgid(struct cifs_sb_info *cifs_sb, void *ptr) { - u32 gid = le32_to_cpu(*(__le32 *)ptr); + u32 gid = get_unaligned_le32(ptr); if (cifs_sb_flags(cifs_sb) & CIFS_MOUNT_OVERR_GID) return cifs_sb->ctx->linux_gid; diff --git a/fs/smb/client/sess.c b/fs/smb/client/sess.c index 7cf7dd104f7c..e095f41b5882 100644 --- a/fs/smb/client/sess.c +++ b/fs/smb/client/sess.c @@ -149,9 +149,9 @@ int cifs_try_adding_channels(struct cifs_ses *ses) int old_chan_count, new_chan_count; int left; int rc = 0; - int tries = 0; + int tries = 0, attempts; size_t iface_weight = 0, iface_min_speed = 0; - struct cifs_server_iface *iface = NULL, *niface = NULL; + struct cifs_server_iface *iface = NULL, *candidate = NULL; struct cifs_server_iface *last_iface = NULL; spin_lock(&ses->chan_lock); @@ -197,67 +197,89 @@ int cifs_try_adding_channels(struct cifs_ses *ses) break; } - if (!iface) - iface = list_first_entry(&ses->iface_list, struct cifs_server_iface, - iface_head); last_iface = list_last_entry(&ses->iface_list, struct cifs_server_iface, iface_head); iface_min_speed = last_iface->speed; + spin_unlock(&ses->iface_lock); - list_for_each_entry_safe_from(iface, niface, &ses->iface_list, - iface_head) { - /* do not mix rdma and non-rdma interfaces */ - if (iface->rdma_capable != ses->server->rdma) - continue; - - /* skip ifaces that are unusable */ - if (!iface->is_active || - (is_ses_using_iface(ses, iface) && - !iface->rss_capable)) - continue; + attempts = 0; + while (left > 0) { + spin_lock(&ses->iface_lock); - /* check if we already allocated enough channels */ - iface_weight = iface->speed / iface_min_speed; + /* + * iface_lock must be dropped while opening a channel, + * and a concurrent interface refresh may remove and + * free entries during that window, so no list entry + * may be kept across it without a reference. Scan + * the list from the beginning each time and only pass + * a referenced candidate to cifs_ses_add_channel(); + * weight_fulfilled tracks the progress so that no + * iface is selected beyond its weight. + */ + candidate = NULL; + list_for_each_entry(iface, &ses->iface_list, iface_head) { + /* do not mix rdma and non-rdma interfaces */ + if (iface->rdma_capable != ses->server->rdma) + continue; + + /* skip ifaces that are unusable */ + if (!iface->is_active || + (is_ses_using_iface(ses, iface) && + !iface->rss_capable)) + continue; + + /* check if we already allocated enough channels */ + iface_weight = iface->speed / iface_min_speed; + + if (iface->weight_fulfilled >= iface_weight) + continue; + + /* take ref before unlock */ + kref_get(&iface->refcount); + candidate = iface; + break; + } - if (iface->weight_fulfilled >= iface_weight) - continue; + if (!candidate) { + /* no usable iface. reset weight_fulfilled and start over */ + list_for_each_entry(iface, &ses->iface_list, iface_head) + iface->weight_fulfilled = 0; + spin_unlock(&ses->iface_lock); + break; + } - /* take ref before unlock */ - kref_get(&iface->refcount); + attempts++; + if (attempts > 3 * ses->chan_max) { + kref_put(&candidate->refcount, release_iface); + spin_unlock(&ses->iface_lock); + break; + } spin_unlock(&ses->iface_lock); - rc = cifs_ses_add_channel(ses, iface); + rc = cifs_ses_add_channel(ses, candidate); spin_lock(&ses->iface_lock); if (rc) { cifs_dbg(VFS, "failed to open extra channel on iface:%pIS rc=%d\n", - &iface->sockaddr, + &candidate->sockaddr, rc); /* failure to add chan should increase weight */ - iface->weight_fulfilled++; - kref_put(&iface->refcount, release_iface); + candidate->weight_fulfilled++; + kref_put(&candidate->refcount, release_iface); + spin_unlock(&ses->iface_lock); continue; } - iface->num_channels++; - iface->weight_fulfilled++; + candidate->num_channels++; + candidate->weight_fulfilled++; cifs_info("successfully opened new channel on iface:%pIS\n", - &iface->sockaddr); - break; - } - - /* reached end of list. reset weight_fulfilled and start over */ - if (list_entry_is_head(iface, &ses->iface_list, iface_head)) { - list_for_each_entry(iface, &ses->iface_list, iface_head) - iface->weight_fulfilled = 0; + &candidate->sockaddr); spin_unlock(&ses->iface_lock); - iface = NULL; - continue; - } - spin_unlock(&ses->iface_lock); - left--; - new_chan_count++; + left--; + new_chan_count++; + break; + } } return new_chan_count - old_chan_count; diff --git a/fs/smb/client/smb2inode.c b/fs/smb/client/smb2inode.c index 96063e355186..13fe8e3b48f3 100644 --- a/fs/smb/client/smb2inode.c +++ b/fs/smb/client/smb2inode.c @@ -77,6 +77,17 @@ static int parse_posix_sids(struct cifs_open_info_data *data, sidsbuf = (u8 *)qi + le16_to_cpu(qi->OutputBufferOffset) + qi_len; sidsbuf_end = sidsbuf + out_len - qi_len; + if (sidsbuf_end < sidsbuf) { + cifs_dbg(VFS, "%s: server-supplied out_len %u caused pointer wraparound\n", + __func__, out_len); + return -EINVAL; + } + if (sidsbuf_end > (u8 *)rsp_iov->iov_base + rsp_iov->iov_len) { + cifs_dbg(VFS, "%s: server-supplied out_len %u overruns iov by %td bytes\n", + __func__, out_len, + sidsbuf_end - ((u8 *)rsp_iov->iov_base + rsp_iov->iov_len)); + return -EINVAL; + } owner_len = posix_info_sid_size(sidsbuf, sidsbuf_end); if (owner_len == -1) diff --git a/fs/smb/client/smb2misc.c b/fs/smb/client/smb2misc.c index 9068175e57cd..0cfe60ae42c3 100644 --- a/fs/smb/client/smb2misc.c +++ b/fs/smb/client/smb2misc.c @@ -85,6 +85,36 @@ static const __le16 smb2_rsp_struct_sizes[NUMBER_OF_SMB2_COMMANDS] = { /* SMB2_OPLOCK_BREAK */ cpu_to_le16(24) }; +/* + * Minimum received PDU size for commands whose response carries a + * variable-length data area. A non-zero entry marks the command as + * having one, and gives the length smb2_check_message() requires + * before smb2_get_data_area_len() reads the offset and length fields + * out of the fixed response struct. + */ +static const size_t smb2_min_pdu_len[NUMBER_OF_SMB2_COMMANDS] = { + /* SMB2_NEGOTIATE */ sizeof(struct smb2_negotiate_rsp), + /* SMB2_SESSION_SETUP */ sizeof(struct smb2_sess_setup_rsp), + /* SMB2_LOGOFF */ 0, + /* SMB2_TREE_CONNECT */ 0, + /* SMB2_TREE_DISCONNECT */ 0, + /* SMB2_CREATE */ sizeof(struct smb2_create_rsp), + /* SMB2_CLOSE */ 0, + /* SMB2_FLUSH */ 0, + /* SMB2_READ */ sizeof(struct smb2_read_rsp), + /* SMB2_WRITE */ 0, + /* SMB2_LOCK */ 0, + /* SMB2_IOCTL */ sizeof(struct smb2_ioctl_rsp), + /* SMB2_CANCEL */ 0, + /* SMB2_ECHO */ 0, + /* SMB2_QUERY_DIRECTORY */ sizeof(struct smb2_query_directory_rsp), + /* SMB2_CHANGE_NOTIFY */ sizeof(struct smb2_change_notify_rsp), + /* SMB2_QUERY_INFO */ sizeof(struct smb2_query_info_rsp), + /* SMB2_SET_INFO */ 0, + /* SMB2_OPLOCK_BREAK */ 0, +}; + +#define smb2_has_data_area(cmd) (smb2_min_pdu_len[cmd] != 0) #define SMB311_NEGPROT_BASE_SIZE (sizeof(struct smb2_hdr) + sizeof(struct smb2_negotiate_rsp)) static __u32 get_neg_ctxt_len(struct smb2_hdr *hdr, __u32 len, @@ -233,6 +263,16 @@ smb2_check_message(char *buf, unsigned int pdu_len, unsigned int len, } } + if ((shdr->Status == STATUS_SUCCESS || + shdr->Status == STATUS_MORE_PROCESSING_REQUIRED || + pdu->StructureSize2 != SMB2_ERROR_STRUCTURE_SIZE2_LE) && + smb2_has_data_area(command) && + len < smb2_min_pdu_len[command]) { + cifs_server_dbg(VFS, "SMB2 command %d response too short: %u < %zu\n", + command, len, smb2_min_pdu_len[command]); + return 1; + } + have_data = false; data_area_overlap = false; calc_len = __smb2_calc_size(buf, &have_data, &data_area_overlap); @@ -299,33 +339,6 @@ smb2_check_message(char *buf, unsigned int pdu_len, unsigned int len, } /* - * The size of the variable area depends on the offset and length fields - * located in different fields for various SMB2 responses. SMB2 responses - * with no variable length info, show an offset of zero for the offset field. - */ -static const bool has_smb2_data_area[NUMBER_OF_SMB2_COMMANDS] = { - /* SMB2_NEGOTIATE */ true, - /* SMB2_SESSION_SETUP */ true, - /* SMB2_LOGOFF */ false, - /* SMB2_TREE_CONNECT */ false, - /* SMB2_TREE_DISCONNECT */ false, - /* SMB2_CREATE */ true, - /* SMB2_CLOSE */ false, - /* SMB2_FLUSH */ false, - /* SMB2_READ */ true, - /* SMB2_WRITE */ false, - /* SMB2_LOCK */ false, - /* SMB2_IOCTL */ true, - /* SMB2_CANCEL */ false, /* BB CHECK this not listed in documentation */ - /* SMB2_ECHO */ false, - /* SMB2_QUERY_DIRECTORY */ true, - /* SMB2_CHANGE_NOTIFY */ true, - /* SMB2_QUERY_INFO */ true, - /* SMB2_SET_INFO */ false, - /* SMB2_OPLOCK_BREAK */ false -}; - -/* * Returns the pointer to the beginning of the data area. Length of the data * area and the offset to it (from the beginning of the smb are also returned. */ @@ -451,7 +464,7 @@ __smb2_calc_size(void *buf, bool *have_data, bool *data_area_overlap) */ len += le16_to_cpu(pdu->StructureSize2); - if (has_smb2_data_area[le16_to_cpu(shdr->Command)] == false) + if (!smb2_has_data_area(le16_to_cpu(shdr->Command))) goto calc_size_exit; smb2_get_data_area_len(&offset, &data_length, shdr); diff --git a/fs/smb/client/smb2ops.c b/fs/smb/client/smb2ops.c index cb4fd09f996e..3464470d3297 100644 --- a/fs/smb/client/smb2ops.c +++ b/fs/smb/client/smb2ops.c @@ -785,9 +785,9 @@ next_iface: break; } /* Validate that Next doesn't point beyond the buffer */ - if (next > bytes_left) { - cifs_dbg(VFS, "%s: invalid Next pointer %zu > %zd\n", - __func__, next, bytes_left); + if (next < sizeof(*p) || next > bytes_left) { + cifs_dbg(VFS, "%s: invalid Next pointer %zu out of range [%zu, %zd]\n", + __func__, next, sizeof(*p), bytes_left); rc = -EINVAL; goto out; } @@ -1053,8 +1053,9 @@ move_smb2_ea_to_cifs(char *dst, size_t dst_size, char *name, *value; size_t buf_size = dst_size; size_t name_len, value_len, user_name_len; + u32 next_off; - while (src_size > 0) { + while (src_size >= sizeof(*src)) { name_len = (size_t)src->ea_name_length; value_len = (size_t)le16_to_cpu(src->ea_value_length); @@ -1110,14 +1111,22 @@ move_smb2_ea_to_cifs(char *dst, size_t dst_size, if (!src->next_entry_offset) break; - if (src_size < le32_to_cpu(src->next_entry_offset)) { - /* stop before overrun buffer */ - rc = -ERANGE; - break; + next_off = le32_to_cpu(src->next_entry_offset); + if (next_off < sizeof(*src) || src_size < next_off) { + cifs_dbg(FYI, "EA next_entry_offset %u out of range [%zu, %zu]\n", + next_off, sizeof(*src), src_size); + rc = smb_EIO2(smb_eio_trace_ea_next_offset, + next_off, src_size); + goto out; + } + src_size -= next_off; + src = (void *)((char *)src + next_off); + if (src_size > 0 && src_size < sizeof(*src)) { + cifs_dbg(FYI, "EA next_entry_offset %u left truncated entry (%zu bytes)\n", + next_off, src_size); + rc = smb_EIO2(smb_eio_trace_ea_next_offset, next_off, src_size); + goto out; } - src_size -= le32_to_cpu(src->next_entry_offset); - src = (void *)((char *)src + - le32_to_cpu(src->next_entry_offset)); } /* didn't find the named attribute */ @@ -2454,8 +2463,14 @@ smb3_enum_snapshots(const unsigned int xid, struct cifs_tcon *tcon, * and retry the ioctl again with larger array size sufficient * to hold all of the snapshot GMT tokens on the second try. */ - if (snapshot_in.snapshot_array_size < GMT_TOKEN_SIZE) + if (snapshot_in.snapshot_array_size < GMT_TOKEN_SIZE) { + if (ret_data_len < sizeof(struct smb_snapshot_array)) { + rc = -EIO; + kfree(retbuf); + return rc; + } ret_data_len = sizeof(struct smb_snapshot_array); + } /* * We return struct SRV_SNAPSHOT_ARRAY, followed by @@ -5365,11 +5380,13 @@ receive_encrypted_standard(struct TCP_Server_Info *server, length = decrypt_raw_data(server, buf, buf_size, NULL, false); if (length) return length; + pdu_length = buf_size; next_is_large = server->large_buf; one_more: shdr = (struct smb2_hdr *)buf; next_cmd = le32_to_cpu(shdr->NextCommand); + server->total_read = next_cmd ? next_cmd : pdu_length; if (*num_mids >= MAX_COMPOUND) { cifs_server_dbg(VFS, "too many PDUs in compound\n"); @@ -5377,8 +5394,15 @@ one_more: } if (next_cmd) { - if (WARN_ON_ONCE(next_cmd > pdu_length)) + if (next_cmd < MID_HEADER_SIZE(server) || + next_cmd > pdu_length || + pdu_length - next_cmd < MID_HEADER_SIZE(server)) { + unsigned int max_next = pdu_length > (unsigned int)MID_HEADER_SIZE(server) ? + pdu_length - (unsigned int)MID_HEADER_SIZE(server) : 0; + cifs_server_dbg(VFS, "invalid NextCommand offset %u out of range [%zu, %u]\n", + next_cmd, MID_HEADER_SIZE(server), max_next); return -1; + } if (next_is_large) next_buffer = (char *)cifs_buf_get(); else @@ -5414,6 +5438,7 @@ one_more: server->bigbuf = buf = next_buffer; else server->smallbuf = buf = next_buffer; + next_buffer = NULL; goto one_more; } else if (ret != 0) { /* diff --git a/fs/smb/client/smb2pdu.c b/fs/smb/client/smb2pdu.c index dea05aeb53a1..880ce12f50c4 100644 --- a/fs/smb/client/smb2pdu.c +++ b/fs/smb/client/smb2pdu.c @@ -189,18 +189,19 @@ cifs_chan_skip_or_disable(struct cifs_ses *ses, spin_unlock(&ses->chan_lock); /* - * the above reference of server by channel - * needs to be dropped without holding chan_lock - * as cifs_put_tcp_session takes a higher lock - * i.e. cifs_tcp_ses_lock + * signal the channel and its primary server to + * reconnect before dropping the above reference of + * server by channel, which is done without holding + * chan_lock as cifs_put_tcp_session takes a higher + * lock i.e. cifs_tcp_ses_lock */ - cifs_put_tcp_session(server, from_reconnect); - cifs_signal_cifsd_for_reconnect(server, false); /* mark primary server as needing reconnect */ pserver = server->primary_server; cifs_signal_cifsd_for_reconnect(pserver, false); + + cifs_put_tcp_session(server, from_reconnect); skip_terminate: return -EHOSTDOWN; } diff --git a/fs/smb/client/trace.h b/fs/smb/client/trace.h index b442cccd1530..bb8d0197cb54 100644 --- a/fs/smb/client/trace.h +++ b/fs/smb/client/trace.h @@ -27,6 +27,7 @@ EM(smb_eio_trace_copychunk_overcopy_c, "copychunk_overcopy_c") \ EM(smb_eio_trace_create_rsp_too_small, "create_rsp_too_small") \ EM(smb_eio_trace_dfsref_no_rsp, "dfsref_no_rsp") \ + EM(smb_eio_trace_ea_next_offset, "ea_next_offset") \ EM(smb_eio_trace_ea_overrun, "ea_overrun") \ EM(smb_eio_trace_extract_will_pin, "extract_will_pin") \ EM(smb_eio_trace_forced_shutdown, "forced_shutdown") \ diff --git a/fs/smb/server/connection.c b/fs/smb/server/connection.c index 4cb92d6599ee..d211861ff86f 100644 --- a/fs/smb/server/connection.c +++ b/fs/smb/server/connection.c @@ -22,6 +22,8 @@ static DEFINE_MUTEX(init_lock); static struct ksmbd_conn_ops default_conn_ops; +static struct delayed_work session_expiration_work; +static bool stopping_session_expiration_work; DEFINE_HASHTABLE(conn_list, CONN_HASH_BITS); DECLARE_RWSEM(conn_list_lock); @@ -158,18 +160,28 @@ static void delete_proc_clients(void) {} static struct workqueue_struct *ksmbd_conn_wq; +static void ksmbd_session_expiration_worker(struct work_struct *work); + int ksmbd_conn_wq_init(void) { ksmbd_conn_wq = alloc_workqueue("ksmbd-conn-release", WQ_UNBOUND | WQ_MEM_RECLAIM, 0); if (!ksmbd_conn_wq) return -ENOMEM; + + WRITE_ONCE(stopping_session_expiration_work, false); + INIT_DELAYED_WORK(&session_expiration_work, + ksmbd_session_expiration_worker); + queue_delayed_work(ksmbd_conn_wq, &session_expiration_work, + KSMBD_SESSION_EXPIRATION_INTERVAL); return 0; } void ksmbd_conn_wq_destroy(void) { if (ksmbd_conn_wq) { + WRITE_ONCE(stopping_session_expiration_work, true); + cancel_delayed_work_sync(&session_expiration_work); destroy_workqueue(ksmbd_conn_wq); ksmbd_conn_wq = NULL; } @@ -279,6 +291,7 @@ struct ksmbd_conn *ksmbd_conn_alloc(void) return NULL; conn->need_neg = true; + conn->creation_time = jiffies; ksmbd_conn_set_new(conn); conn->local_nls = load_nls("utf8"); if (!conn->local_nls) @@ -588,6 +601,20 @@ bool ksmbd_conn_alive(struct ksmbd_conn *conn) if (kthread_should_stop()) return false; + /* + * Stale connections that have not completed NEGOTIATE and SESSION_SETUP + * must be disconnected. Do not race a request that is currently + * completing authentication. + */ + if (!atomic_read(&conn->req_running) && + time_after(jiffies, conn->creation_time + + KSMBD_UNAUTHENTICATED_CONN_TIMEOUT) && + (READ_ONCE(conn->need_neg) || + !ksmbd_conn_has_valid_or_expired_session(conn))) { + ksmbd_debug(CONN, "Connection setup timed out\n"); + return false; + } + if (atomic_read(&conn->stats.open_files_count) > 0) return true; @@ -605,6 +632,51 @@ bool ksmbd_conn_alive(struct ksmbd_conn *conn) return true; } +static void ksmbd_session_expiration_worker(struct work_struct *work) +{ + struct ksmbd_conn *conn, *target; + int bkt; + + if (!ksmbd_server_running()) + goto reschedule; + + ksmbd_expire_sessions(); + + /* + * An old connection without a Valid or Expired session must be + * disconnected. Process one connection at a time without holding + * conn_list_lock across transport shutdown. + */ +again: + target = NULL; + down_read(&conn_list_lock); + hash_for_each(conn_list, bkt, conn, hlist) { + if (ksmbd_conn_exiting(conn) || ksmbd_conn_releasing(conn) || + atomic_read(&conn->req_running) || + time_before_eq(jiffies, conn->creation_time + + KSMBD_UNAUTHENTICATED_CONN_TIMEOUT) || + (!READ_ONCE(conn->need_neg) && + ksmbd_conn_has_valid_or_expired_session(conn))) + continue; + + target = ksmbd_conn_get(conn); + break; + } + up_read(&conn_list_lock); + + if (target) { + ksmbd_debug(CONN, "Connection setup timed out\n"); + ksmbd_conn_abort(target); + ksmbd_conn_put(target); + goto again; + } + +reschedule: + if (!READ_ONCE(stopping_session_expiration_work)) + queue_delayed_work(ksmbd_conn_wq, &session_expiration_work, + KSMBD_SESSION_EXPIRATION_INTERVAL); +} + /* "+2" for BCC field (ByteCount, 2 bytes) */ #define SMB1_MIN_SUPPORTED_PDU_SIZE (sizeof(struct smb_hdr) + 2) #define SMB2_MIN_SUPPORTED_PDU_SIZE (sizeof(struct smb2_pdu)) diff --git a/fs/smb/server/connection.h b/fs/smb/server/connection.h index 63484c8efbbd..371f17b4f02a 100644 --- a/fs/smb/server/connection.h +++ b/fs/smb/server/connection.h @@ -77,6 +77,7 @@ struct ksmbd_conn { struct rw_semaphore session_lock; /* smb session 1 per user */ struct xarray sessions; + unsigned long creation_time; unsigned long last_active; /* How many request are running currently */ atomic_t req_running; @@ -192,6 +193,8 @@ struct ksmbd_transport { #define KSMBD_TCP_RECV_TIMEOUT (7 * HZ) #define KSMBD_TCP_SEND_TIMEOUT (5 * HZ) +#define KSMBD_SESSION_EXPIRATION_INTERVAL (5 * HZ) +#define KSMBD_UNAUTHENTICATED_CONN_TIMEOUT (45 * HZ) #define KSMBD_TCP_PEER_SOCKADDR(c) ((struct sockaddr *)&((c)->peer_addr)) #define CONN_HASH_BITS 12 diff --git a/fs/smb/server/mgmt/user_session.c b/fs/smb/server/mgmt/user_session.c index 2eb8f730e99e..44dc3f800cd4 100644 --- a/fs/smb/server/mgmt/user_session.c +++ b/fs/smb/server/mgmt/user_session.c @@ -22,6 +22,7 @@ static DEFINE_IDA(session_ida); #define SESSION_HASH_BITS 12 +#define KSMBD_MAX_PENDING_SESSIONS 1 static DEFINE_HASHTABLE(sessions_table, SESSION_HASH_BITS); static DECLARE_RWSEM(sessions_table_lock); @@ -432,26 +433,31 @@ struct ksmbd_session *__session_lookup(unsigned long long id) return NULL; } -static void ksmbd_expire_session(struct ksmbd_conn *conn) +static bool ksmbd_too_many_session_setups(struct ksmbd_conn *conn) { unsigned long id; struct ksmbd_session *sess; + unsigned int pending = 0; down_write(&sessions_table_lock); down_write(&conn->session_lock); xa_for_each(&conn->sessions, id, sess) { + if (READ_ONCE(sess->state) != SMB2_SESSION_IN_PROGRESS) + continue; + if (atomic_read(&sess->refcnt) <= 1 && - (sess->state != SMB2_SESSION_VALID || - time_after(jiffies, - sess->last_active + SMB2_SESSION_TIMEOUT))) { + time_after(jiffies, sess->last_active + + KSMBD_UNAUTHENTICATED_CONN_TIMEOUT)) { xa_erase(&conn->sessions, sess->id); ksmbd_session_remove_from_table(sess); ksmbd_session_destroy(sess); continue; } + pending++; } up_write(&conn->session_lock); up_write(&sessions_table_lock); + return pending >= KSMBD_MAX_PENDING_SESSIONS; } int ksmbd_session_register(struct ksmbd_conn *conn, @@ -461,9 +467,12 @@ int ksmbd_session_register(struct ksmbd_conn *conn, sess->dialect = conn->dialect; memcpy(sess->ClientGUID, conn->ClientGUID, SMB2_CLIENT_GUID_SIZE); - ksmbd_expire_session(conn); - ret = xa_err(xa_store(&conn->sessions, sess->id, sess, - KSMBD_DEFAULT_GFP)); + /* Bound abandoned SessionId-zero authentication exchanges. */ + if (ksmbd_too_many_session_setups(conn)) + ret = -ENOSPC; + else + ret = xa_err(xa_store(&conn->sessions, sess->id, sess, + KSMBD_DEFAULT_GFP)); if (ret) { down_write(&sessions_table_lock); ksmbd_session_remove_from_table(sess); @@ -474,6 +483,105 @@ int ksmbd_session_register(struct ksmbd_conn *conn, return ret; } +void ksmbd_session_unregister(struct ksmbd_conn *conn, + struct ksmbd_session *sess) +{ + struct ksmbd_conn *session_conns[KSMBD_MAX_CHANNELS]; + struct channel *chann; + unsigned long index; + unsigned int nr_conns = 0, i; + bool removed = false; + + down_write(&sessions_table_lock); + if (!hlist_unhashed(&sess->hlist)) { + /* Keep each channel connection stable under sessions_table_lock. */ + down_read(&sess->chann_lock); + xa_for_each(&sess->ksmbd_chann_list, index, chann) { + if (nr_conns == ARRAY_SIZE(session_conns)) + break; + session_conns[nr_conns++] = chann->conn; + } + up_read(&sess->chann_lock); + + ksmbd_session_remove_from_table(sess); + removed = true; + } + + down_write(&conn->session_lock); + if (xa_load(&conn->sessions, sess->id) == sess) + xa_erase(&conn->sessions, sess->id); + up_write(&conn->session_lock); + for (i = 0; i < nr_conns; i++) { + if (session_conns[i] == conn) + continue; + down_write(&session_conns[i]->session_lock); + if (xa_load(&session_conns[i]->sessions, sess->id) == sess) + xa_erase(&session_conns[i]->sessions, sess->id); + up_write(&session_conns[i]->session_lock); + } + up_write(&sessions_table_lock); + + if (removed) + ksmbd_user_session_put(sess); +} + +bool ksmbd_conn_has_valid_or_expired_session(struct ksmbd_conn *conn) +{ + struct ksmbd_session *sess; + unsigned long id; + int state, bkt; + bool found = false; + + down_read(&conn->session_lock); + xa_for_each(&conn->sessions, id, sess) { + state = READ_ONCE(sess->state); + if (state == SMB2_SESSION_VALID || + state == SMB2_SESSION_EXPIRED) { + found = true; + break; + } + } + up_read(&conn->session_lock); + if (found) + return true; + + /* A session bound through SMB3 multichannel is not in conn->sessions. */ + down_read(&sessions_table_lock); + hash_for_each(sessions_table, bkt, sess, hlist) { + state = READ_ONCE(sess->state); + if (state != SMB2_SESSION_VALID && + state != SMB2_SESSION_EXPIRED) + continue; + + down_read(&sess->chann_lock); + found = xa_load(&sess->ksmbd_chann_list, (long)conn); + up_read(&sess->chann_lock); + if (found) + break; + } + up_read(&sessions_table_lock); + return found; +} + +void ksmbd_expire_sessions(void) +{ + struct ksmbd_session *sess; + u64 now = ktime_get_real_seconds(); + int bkt; + + down_read(&sessions_table_lock); + hash_for_each(sessions_table, bkt, sess, hlist) { + if (READ_ONCE(sess->state) != SMB2_SESSION_VALID || + !sess->kerberos_expiry || now < sess->kerberos_expiry) + continue; + + if (cmpxchg(&sess->state, SMB2_SESSION_VALID, + SMB2_SESSION_EXPIRED) == SMB2_SESSION_VALID) + ksmbd_counter_inc(KSMBD_COUNTER_SESSION_TIMEOUTS); + } + up_read(&sessions_table_lock); +} + static int ksmbd_chann_del(struct ksmbd_conn *conn, struct ksmbd_session *sess) { struct channel *chann; @@ -488,7 +596,7 @@ static int ksmbd_chann_del(struct ksmbd_conn *conn, struct ksmbd_session *sess) return 0; } -void ksmbd_sessions_deregister(struct ksmbd_conn *conn) +void ksmbd_conn_sessions_cleanup(struct ksmbd_conn *conn) { struct ksmbd_session *sess; unsigned long id; diff --git a/fs/smb/server/mgmt/user_session.h b/fs/smb/server/mgmt/user_session.h index 3e52d4cc1324..217258551d6d 100644 --- a/fs/smb/server/mgmt/user_session.h +++ b/fs/smb/server/mgmt/user_session.h @@ -72,6 +72,8 @@ struct ksmbd_session { struct rw_semaphore rpc_lock; }; +#define KSMBD_MAX_CHANNELS 32 + static inline int test_session_flag(struct ksmbd_session *sess, int bit) { return sess->flags & bit; @@ -98,7 +100,11 @@ bool is_ksmbd_session_in_connection(struct ksmbd_conn *conn, unsigned long long id); int ksmbd_session_register(struct ksmbd_conn *conn, struct ksmbd_session *sess); -void ksmbd_sessions_deregister(struct ksmbd_conn *conn); +void ksmbd_session_unregister(struct ksmbd_conn *conn, + struct ksmbd_session *sess); +void ksmbd_conn_sessions_cleanup(struct ksmbd_conn *conn); +bool ksmbd_conn_has_valid_or_expired_session(struct ksmbd_conn *conn); +void ksmbd_expire_sessions(void); struct ksmbd_session *__session_lookup(unsigned long long id); struct ksmbd_session *ksmbd_session_lookup_all(struct ksmbd_conn *conn, unsigned long long id); diff --git a/fs/smb/server/proc.c b/fs/smb/server/proc.c index 826353ed0553..19f0f2cfbf54 100644 --- a/fs/smb/server/proc.c +++ b/fs/smb/server/proc.c @@ -178,6 +178,8 @@ static int proc_show_ksmbd_stats(struct seq_file *m, void *v) proc_show_runtime_totals(m); seq_printf(m, "sessions:\t%lld\n", ksmbd_counter_sum(KSMBD_COUNTER_SESSIONS)); + seq_printf(m, "session_timeouts:\t%lld\n", + ksmbd_counter_sum(KSMBD_COUNTER_SESSION_TIMEOUTS)); seq_printf(m, "tree_connects:\t%lld\n", ksmbd_counter_sum(KSMBD_COUNTER_TREE_CONNS)); seq_printf(m, "requests:\t%lld\n", diff --git a/fs/smb/server/server.c b/fs/smb/server/server.c index 0069d4e6a60a..0827c8c51006 100644 --- a/fs/smb/server/server.c +++ b/fs/smb/server/server.c @@ -414,7 +414,7 @@ static int ksmbd_server_process_request(struct ksmbd_conn *conn) static int ksmbd_server_terminate_conn(struct ksmbd_conn *conn) { - ksmbd_sessions_deregister(conn); + ksmbd_conn_sessions_cleanup(conn); destroy_lease_table(conn); return 0; } diff --git a/fs/smb/server/smb2pdu.c b/fs/smb/server/smb2pdu.c index b7ce67094626..6b8809f67b92 100644 --- a/fs/smb/server/smb2pdu.c +++ b/fs/smb/server/smb2pdu.c @@ -85,8 +85,6 @@ struct channel *lookup_chann_list(struct ksmbd_session *sess, struct ksmbd_conn return chann; } -#define KSMBD_MAX_CHANNELS 32 - static int register_session_channel(struct ksmbd_session *sess, struct ksmbd_conn *conn, const char *sess_key) @@ -933,8 +931,14 @@ static bool smb2_session_expired_cmd_allowed(struct ksmbd_work *work, static bool smb2_session_kerberos_expired(struct ksmbd_session *sess) { - return sess->kerberos_expiry && - ktime_get_real_seconds() >= sess->kerberos_expiry; + if (!sess->kerberos_expiry || + ktime_get_real_seconds() < sess->kerberos_expiry) + return false; + + if (cmpxchg(&sess->state, SMB2_SESSION_VALID, + SMB2_SESSION_EXPIRED) == SMB2_SESSION_VALID) + ksmbd_counter_inc(KSMBD_COUNTER_SESSION_TIMEOUTS); + return true; } /** @@ -969,9 +973,8 @@ int smb2_check_user_session(struct ksmbd_work *work) if (!work->next_smb2_rcv_hdr_off && sess_id) work->sess = ksmbd_session_lookup_all_states(conn, sess_id); if (work->sess) { - if (smb2_session_kerberos_expired(work->sess)) { - work->sess->state = SMB2_SESSION_EXPIRED; - } else if (work->sess->state != SMB2_SESSION_VALID) { + if (!smb2_session_kerberos_expired(work->sess) && + work->sess->state != SMB2_SESSION_VALID) { ksmbd_user_session_put(work->sess); work->sess = NULL; } @@ -996,8 +999,7 @@ int smb2_check_user_session(struct ksmbd_work *work) sess_id, work->sess->id); return -EINVAL; } - if (smb2_session_kerberos_expired(work->sess)) - work->sess->state = SMB2_SESSION_EXPIRED; + smb2_session_kerberos_expired(work->sess); if (work->sess->state != SMB2_SESSION_VALID) { pr_err("compound request on a non-valid session (state %d)\n", work->sess->state); @@ -1014,7 +1016,6 @@ int smb2_check_user_session(struct ksmbd_work *work) work->sess = ksmbd_session_lookup_all_states(conn, sess_id); if (work->sess) { if (smb2_session_kerberos_expired(work->sess)) { - work->sess->state = SMB2_SESSION_EXPIRED; return smb2_session_expired_cmd_allowed(work, cmd) ? 1 : -EKEYEXPIRED; } @@ -2436,7 +2437,7 @@ int smb2_sess_setup(struct ksmbd_work *work) struct ksmbd_conn *conn = work->conn; struct smb2_sess_setup_req *req; struct smb2_sess_setup_rsp *rsp; - struct ksmbd_session *sess; + struct ksmbd_session *sess = NULL; struct negotiate_message *negblob; unsigned int negblob_len, negblob_off; int rc = 0; @@ -2592,6 +2593,9 @@ int smb2_sess_setup(struct ksmbd_work *work) goto out_err; } + if (work->session_setup_reauth) + WRITE_ONCE(sess->state, SMB2_SESSION_IN_PROGRESS); + conn->binding = false; } work->sess = sess; @@ -2703,6 +2707,14 @@ out_err: } if (rc < 0) { + bool setup_in_progress = sess && + READ_ONCE(sess->state) == SMB2_SESSION_IN_PROGRESS && + !(req->Flags & SMB2_SESSION_REQ_FLAG_BINDING); + + /* Authentication errors must not leave the new session published. */ + if (setup_in_progress) + ksmbd_session_unregister(conn, sess); + if (sess && conn->dialect == SMB311_PROT_ID && (req->Flags & SMB2_SESSION_REQ_FLAG_BINDING)) { struct preauth_session *preauth_sess; @@ -2736,7 +2748,8 @@ out_err: * For binding requests, session belongs to another * connection. Do not expire it. */ - if (!(req->Flags & SMB2_SESSION_REQ_FLAG_BINDING)) { + if (!(req->Flags & SMB2_SESSION_REQ_FLAG_BINDING) && + !setup_in_progress) { sess->last_active = jiffies; sess->kerberos_expiry = 0; sess->state = SMB2_SESSION_EXPIRED; @@ -6262,21 +6275,18 @@ err_out2: * @reqOutputBufferLength: max buffer length expected in command response * @fixed_len: minimum fixed response length * @rsp: query info response buffer contains output buffer length - * @rsp_org: base response buffer pointer in case of chained response * * Return: 0 on success, otherwise error */ static int buffer_check_err(int reqOutputBufferLength, unsigned int fixed_len, - struct smb2_query_info_rsp *rsp, - void *rsp_org) + struct smb2_query_info_rsp *rsp) { unsigned int output_len = le32_to_cpu(rsp->OutputBufferLength); if (reqOutputBufferLength < fixed_len) { pr_err("Invalid Buffer Size Requested\n"); rsp->hdr.Status = STATUS_INFO_LENGTH_MISMATCH; - *(__be32 *)rsp_org = cpu_to_be32(sizeof(struct smb2_hdr)); return -EINVAL; } @@ -6287,8 +6297,7 @@ static int buffer_check_err(int reqOutputBufferLength, return 0; } -static void get_standard_info_pipe(struct smb2_query_info_rsp *rsp, - void *rsp_org) +static void get_standard_info_pipe(struct smb2_query_info_rsp *rsp) { struct smb2_file_standard_info *sinfo; @@ -6303,8 +6312,7 @@ static void get_standard_info_pipe(struct smb2_query_info_rsp *rsp, cpu_to_le32(sizeof(struct smb2_file_standard_info)); } -static void get_internal_info_pipe(struct smb2_query_info_rsp *rsp, u64 num, - void *rsp_org) +static void get_internal_info_pipe(struct smb2_query_info_rsp *rsp, u64 num) { struct smb2_file_internal_info *file_info; @@ -6318,8 +6326,7 @@ static void get_internal_info_pipe(struct smb2_query_info_rsp *rsp, u64 num, static int smb2_get_info_file_pipe(struct ksmbd_session *sess, struct smb2_query_info_req *req, - struct smb2_query_info_rsp *rsp, - void *rsp_org) + struct smb2_query_info_rsp *rsp) { u64 id; int rc; @@ -6344,16 +6351,16 @@ static int smb2_get_info_file_pipe(struct ksmbd_session *sess, switch (req->FileInfoClass) { case FILE_STANDARD_INFORMATION: - get_standard_info_pipe(rsp, rsp_org); + get_standard_info_pipe(rsp); rc = buffer_check_err(le32_to_cpu(req->OutputBufferLength), le32_to_cpu(rsp->OutputBufferLength), - rsp, rsp_org); + rsp); break; case FILE_INTERNAL_INFORMATION: - get_internal_info_pipe(rsp, id, rsp_org); + get_internal_info_pipe(rsp, id); rc = buffer_check_err(le32_to_cpu(req->OutputBufferLength), le32_to_cpu(rsp->OutputBufferLength), - rsp, rsp_org); + rsp); break; default: ksmbd_debug(SMB, "smb2_info_file_pipe for %u not supported\n", @@ -7189,8 +7196,7 @@ static int smb2_get_info_file(struct ksmbd_work *work, if (test_share_config_flag(work->tcon->share_conf, KSMBD_SHARE_FLAG_PIPE)) { /* smb2 info file called for pipe */ - rc = smb2_get_info_file_pipe(work->sess, req, rsp, - work->response_buf); + rc = smb2_get_info_file_pipe(work->sess, req, rsp); goto iov_pin_out; } @@ -7299,13 +7305,16 @@ static int smb2_get_info_file(struct ksmbd_work *work, case FILE_ALTERNATE_NAME_INFORMATION: fixed_len = FILE_ALTERNATE_NAME_INFORMATION_SIZE; break; + case FILE_NORMALIZED_NAME_INFORMATION: + fixed_len = FILE_NORMALIZED_NAME_INFORMATION_SIZE; + break; case FILE_STREAM_INFORMATION: fixed_len = FILE_STREAM_INFORMATION_SIZE; break; } rc = buffer_check_err(le32_to_cpu(req->OutputBufferLength), fixed_len, - rsp, work->response_buf); + rsp); } ksmbd_fd_put(work, fp); @@ -7576,7 +7585,7 @@ static int smb2_get_info_filesystem(struct ksmbd_work *work, } rc = buffer_check_err(le32_to_cpu(req->OutputBufferLength), fixed_len, - rsp, work->response_buf); + rsp); path_put(&path); if (!rc) @@ -7690,7 +7699,7 @@ release_acl: rsp->OutputBufferLength = cpu_to_le32(secdesclen); rc = buffer_check_err(le32_to_cpu(req->OutputBufferLength), le32_to_cpu(rsp->OutputBufferLength), - rsp, work->response_buf); + rsp); if (rc) goto err_out; diff --git a/fs/smb/server/smb2pdu.h b/fs/smb/server/smb2pdu.h index 3f08d1ca5a38..ca8e27f7b712 100644 --- a/fs/smb/server/smb2pdu.h +++ b/fs/smb/server/smb2pdu.h @@ -61,8 +61,6 @@ struct preauth_integrity_info { #define SMB2_SESSION_IN_PROGRESS BIT(0) #define SMB2_SESSION_VALID BIT(1) -#define SMB2_SESSION_TIMEOUT (10 * HZ) - /* Apple Defined Contexts */ #define SMB2_CREATE_AAPL "AAPL" @@ -214,6 +212,7 @@ struct file_sparse { #define FILE_ALLOCATION_INFORMATION_SIZE 19 #define FILE_END_OF_FILE_INFORMATION_SIZE 20 #define FILE_ALTERNATE_NAME_INFORMATION_SIZE 8 +#define FILE_NORMALIZED_NAME_INFORMATION_SIZE 8 #define FILE_STREAM_INFORMATION_SIZE 32 #define FILE_PIPE_INFORMATION_SIZE 23 #define FILE_PIPE_LOCAL_INFORMATION_SIZE 24 diff --git a/fs/smb/server/stats.h b/fs/smb/server/stats.h index bc864efa0d46..8b32b8b4e8be 100644 --- a/fs/smb/server/stats.h +++ b/fs/smb/server/stats.h @@ -15,6 +15,7 @@ enum { KSMBD_COUNTER_SESSIONS = 0, + KSMBD_COUNTER_SESSION_TIMEOUTS, KSMBD_COUNTER_TREE_CONNS, KSMBD_COUNTER_REQUESTS, KSMBD_COUNTER_STATUS_SUCCESS, |