From 3746d62ecf1c872a520c4866118edccb121c44fd Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 23 Jun 2022 11:06:43 -0600 Subject: [PATCH 01/11] io_uring: use original request task for inflight tracking commit 386e4fb6962b9f248a80f8870aea0870ca603e89 upstream. In prior kernels, we did file assignment always at prep time. This meant that req->task == current. But after deferring that assignment and then pushing the inflight tracking back in, we've got the inflight tracking using current when it should in fact now be using req->task. Fixup that error introduced by adding the inflight tracking back after file assignments got modifed. Fixes: 9cae36a094e7 ("io_uring: reinstate the inflight tracking") Signed-off-by: Jens Axboe Signed-off-by: Greg Kroah-Hartman --- fs/io_uring.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/io_uring.c b/fs/io_uring.c index 0ce1587df432..89f24b54fe5e 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -1405,7 +1405,7 @@ static void io_req_track_inflight(struct io_kiocb *req) { if (!(req->flags & REQ_F_INFLIGHT)) { req->flags |= REQ_F_INFLIGHT; - atomic_inc(¤t->io_uring->inflight_tracked); + atomic_inc(&req->task->io_uring->inflight_tracked); } } From c12f0e6126ad223806a365084e86370511654bf1 Mon Sep 17 00:00:00 2001 From: Jens Wiklander Date: Thu, 18 Aug 2022 13:08:59 +0200 Subject: [PATCH 02/11] tee: add overflow check in register_shm_helper() commit 573ae4f13f630d6660008f1974c0a8a29c30e18a upstream. With special lengths supplied by user space, register_shm_helper() has an integer overflow when calculating the number of pages covered by a supplied user space memory region. This causes internal_get_user_pages_fast() a helper function of pin_user_pages_fast() to do a NULL pointer dereference: Unable to handle kernel NULL pointer dereference at virtual address 0000000000000010 Modules linked in: CPU: 1 PID: 173 Comm: optee_example_a Not tainted 5.19.0 #11 Hardware name: QEMU QEMU Virtual Machine, BIOS 0.0.0 02/06/2015 pc : internal_get_user_pages_fast+0x474/0xa80 Call trace: internal_get_user_pages_fast+0x474/0xa80 pin_user_pages_fast+0x24/0x4c register_shm_helper+0x194/0x330 tee_shm_register_user_buf+0x78/0x120 tee_ioctl+0xd0/0x11a0 __arm64_sys_ioctl+0xa8/0xec invoke_syscall+0x48/0x114 Fix this by adding an an explicit call to access_ok() in tee_shm_register_user_buf() to catch an invalid user space address early. Fixes: 033ddf12bcf5 ("tee: add register user memory") Cc: stable@vger.kernel.org Reported-by: Nimish Mishra Reported-by: Anirban Chakraborty Reported-by: Debdeep Mukhopadhyay Suggested-by: Jerome Forissier Signed-off-by: Jens Wiklander Signed-off-by: Linus Torvalds Signed-off-by: Greg Kroah-Hartman --- drivers/tee/tee_shm.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/tee/tee_shm.c b/drivers/tee/tee_shm.c index 499fccba3d74..6e662fb131d5 100644 --- a/drivers/tee/tee_shm.c +++ b/drivers/tee/tee_shm.c @@ -222,6 +222,9 @@ struct tee_shm *tee_shm_register(struct tee_context *ctx, unsigned long addr, goto err; } + if (!access_ok((void __user *)addr, length)) + return ERR_PTR(-EFAULT); + mutex_lock(&teedev->mutex); shm->id = idr_alloc(&teedev->idr, shm, 1, 0, GFP_KERNEL); mutex_unlock(&teedev->mutex); From 5430db94434f9100cbf2284903652c5fb26a5ce1 Mon Sep 17 00:00:00 2001 From: Jamal Hadi Salim Date: Sun, 14 Aug 2022 11:27:58 +0000 Subject: [PATCH 03/11] net_sched: cls_route: disallow handle of 0 commit 02799571714dc5dd6948824b9d080b44a295f695 upstream. Follows up on: https://lore.kernel.org/all/20220809170518.164662-1-cascardo@canonical.com/ handle of 0 implies from/to of universe realm which is not very sensible. Lets see what this patch will do: $sudo tc qdisc add dev $DEV root handle 1:0 prio //lets manufacture a way to insert handle of 0 $sudo tc filter add dev $DEV parent 1:0 protocol ip prio 100 \ route to 0 from 0 classid 1:10 action ok //gets rejected... Error: handle of 0 is not valid. We have an error talking to the kernel, -1 //lets create a legit entry.. sudo tc filter add dev $DEV parent 1:0 protocol ip prio 100 route from 10 \ classid 1:10 action ok //what did the kernel insert? $sudo tc filter ls dev $DEV parent 1:0 filter protocol ip pref 100 route chain 0 filter protocol ip pref 100 route chain 0 fh 0x000a8000 flowid 1:10 from 10 action order 1: gact action pass random type none pass val 0 index 1 ref 1 bind 1 //Lets try to replace that legit entry with a handle of 0 $ sudo tc filter replace dev $DEV parent 1:0 protocol ip prio 100 \ handle 0x000a8000 route to 0 from 0 classid 1:10 action drop Error: Replacing with handle of 0 is invalid. We have an error talking to the kernel, -1 And last, lets run Cascardo's POC: $ ./poc 0 0 -22 -22 -22 Signed-off-by: Jamal Hadi Salim Acked-by: Stephen Hemminger Signed-off-by: David S. Miller Signed-off-by: Greg Kroah-Hartman --- net/sched/cls_route.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/net/sched/cls_route.c b/net/sched/cls_route.c index 3f935cbbaff6..48712bc51bda 100644 --- a/net/sched/cls_route.c +++ b/net/sched/cls_route.c @@ -424,6 +424,11 @@ static int route4_set_parms(struct net *net, struct tcf_proto *tp, return -EINVAL; } + if (!nhandle) { + NL_SET_ERR_MSG(extack, "Replacing with handle of 0 is invalid"); + return -EINVAL; + } + h1 = to_hash(nhandle); b = rtnl_dereference(head->table[h1]); if (!b) { @@ -477,6 +482,11 @@ static int route4_change(struct net *net, struct sk_buff *in_skb, int err; bool new = true; + if (!handle) { + NL_SET_ERR_MSG(extack, "Creating with handle of 0 is invalid"); + return -EINVAL; + } + if (opt == NULL) return handle ? -EINVAL : 0; From c76b216753c9eb2950a091037c9976f389e73529 Mon Sep 17 00:00:00 2001 From: Hyunchul Lee Date: Thu, 28 Jul 2022 23:41:51 +0900 Subject: [PATCH 04/11] ksmbd: prevent out of bound read for SMB2_WRITE commit ac60778b87e45576d7bfdbd6f53df902654e6f09 upstream. OOB read memory can be written to a file, if DataOffset is 0 and Length is too large in SMB2_WRITE request of compound request. To prevent this, when checking the length of the data area of SMB2_WRITE in smb2_get_data_area_len(), let the minimum of DataOffset be the size of SMB2 header + the size of SMB2_WRITE header. This bug can lead an oops looking something like: [ 798.008715] BUG: KASAN: slab-out-of-bounds in copy_page_from_iter_atomic+0xd3d/0x14b0 [ 798.008724] Read of size 252 at addr ffff88800f863e90 by task kworker/0:2/2859 ... [ 798.008754] Call Trace: [ 798.008756] [ 798.008759] dump_stack_lvl+0x49/0x5f [ 798.008764] print_report.cold+0x5e/0x5cf [ 798.008768] ? __filemap_get_folio+0x285/0x6d0 [ 798.008774] ? copy_page_from_iter_atomic+0xd3d/0x14b0 [ 798.008777] kasan_report+0xaa/0x120 [ 798.008781] ? copy_page_from_iter_atomic+0xd3d/0x14b0 [ 798.008784] kasan_check_range+0x100/0x1e0 [ 798.008788] memcpy+0x24/0x60 [ 798.008792] copy_page_from_iter_atomic+0xd3d/0x14b0 [ 798.008795] ? pagecache_get_page+0x53/0x160 [ 798.008799] ? iov_iter_get_pages_alloc+0x1590/0x1590 [ 798.008803] ? ext4_write_begin+0xfc0/0xfc0 [ 798.008807] ? current_time+0x72/0x210 [ 798.008811] generic_perform_write+0x2c8/0x530 [ 798.008816] ? filemap_fdatawrite_wbc+0x180/0x180 [ 798.008820] ? down_write+0xb4/0x120 [ 798.008824] ? down_write_killable+0x130/0x130 [ 798.008829] ext4_buffered_write_iter+0x137/0x2c0 [ 798.008833] ext4_file_write_iter+0x40b/0x1490 [ 798.008837] ? __fsnotify_parent+0x275/0xb20 [ 798.008842] ? __fsnotify_update_child_dentry_flags+0x2c0/0x2c0 [ 798.008846] ? ext4_buffered_write_iter+0x2c0/0x2c0 [ 798.008851] __kernel_write+0x3a1/0xa70 [ 798.008855] ? __x64_sys_preadv2+0x160/0x160 [ 798.008860] ? security_file_permission+0x4a/0xa0 [ 798.008865] kernel_write+0xbb/0x360 [ 798.008869] ksmbd_vfs_write+0x27e/0xb90 [ksmbd] [ 798.008881] ? ksmbd_vfs_read+0x830/0x830 [ksmbd] [ 798.008892] ? _raw_read_unlock+0x2a/0x50 [ 798.008896] smb2_write+0xb45/0x14e0 [ksmbd] [ 798.008909] ? __kasan_check_write+0x14/0x20 [ 798.008912] ? _raw_spin_lock_bh+0xd0/0xe0 [ 798.008916] ? smb2_read+0x15e0/0x15e0 [ksmbd] [ 798.008927] ? memcpy+0x4e/0x60 [ 798.008931] ? _raw_spin_unlock+0x19/0x30 [ 798.008934] ? ksmbd_smb2_check_message+0x16af/0x2350 [ksmbd] [ 798.008946] ? _raw_spin_lock_bh+0xe0/0xe0 [ 798.008950] handle_ksmbd_work+0x30e/0x1020 [ksmbd] [ 798.008962] process_one_work+0x778/0x11c0 [ 798.008966] ? _raw_spin_lock_irq+0x8e/0xe0 [ 798.008970] worker_thread+0x544/0x1180 [ 798.008973] ? __cpuidle_text_end+0x4/0x4 [ 798.008977] kthread+0x282/0x320 [ 798.008982] ? process_one_work+0x11c0/0x11c0 [ 798.008985] ? kthread_complete_and_exit+0x30/0x30 [ 798.008989] ret_from_fork+0x1f/0x30 [ 798.008995] Fixes: e2f34481b24d ("cifsd: add server-side procedures for SMB3") Cc: stable@vger.kernel.org Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-17817 Signed-off-by: Hyunchul Lee Acked-by: Namjae Jeon Signed-off-by: Steve French Signed-off-by: Greg Kroah-Hartman --- fs/ksmbd/smb2misc.c | 7 +++++-- fs/ksmbd/smb2pdu.c | 6 ++---- 2 files changed, 7 insertions(+), 6 deletions(-) diff --git a/fs/ksmbd/smb2misc.c b/fs/ksmbd/smb2misc.c index 66b24b480ebf..b47be71be4c8 100644 --- a/fs/ksmbd/smb2misc.c +++ b/fs/ksmbd/smb2misc.c @@ -132,8 +132,11 @@ static int smb2_get_data_area_len(unsigned int *off, unsigned int *len, *len = le16_to_cpu(((struct smb2_read_req *)hdr)->ReadChannelInfoLength); break; case SMB2_WRITE: - if (((struct smb2_write_req *)hdr)->DataOffset) { - *off = le16_to_cpu(((struct smb2_write_req *)hdr)->DataOffset); + if (((struct smb2_write_req *)hdr)->DataOffset || + ((struct smb2_write_req *)hdr)->Length) { + *off = max_t(unsigned int, + le16_to_cpu(((struct smb2_write_req *)hdr)->DataOffset), + offsetof(struct smb2_write_req, Buffer) - 4); *len = le32_to_cpu(((struct smb2_write_req *)hdr)->Length); break; } diff --git a/fs/ksmbd/smb2pdu.c b/fs/ksmbd/smb2pdu.c index 53f5db40b96e..06552a67a810 100644 --- a/fs/ksmbd/smb2pdu.c +++ b/fs/ksmbd/smb2pdu.c @@ -6471,10 +6471,8 @@ int smb2_write(struct ksmbd_work *work) (offsetof(struct smb2_write_req, Buffer) - 4)) { data_buf = (char *)&req->Buffer[0]; } else { - if ((u64)le16_to_cpu(req->DataOffset) + length > get_rfc1002_len(req)) { - pr_err("invalid write data offset %u, smb_len %u\n", - le16_to_cpu(req->DataOffset), - get_rfc1002_len(req)); + if (le16_to_cpu(req->DataOffset) < + offsetof(struct smb2_write_req, Buffer)) { err = -EINVAL; goto out; } From cb69d4d6f709f87c94afa28ae64c501576692171 Mon Sep 17 00:00:00 2001 From: Namjae Jeon Date: Tue, 2 Aug 2022 07:28:51 +0900 Subject: [PATCH 05/11] ksmbd: fix heap-based overflow in set_ntacl_dacl() commit 8f0541186e9ad1b62accc9519cc2b7a7240272a7 upstream. The testcase use SMB2_SET_INFO_HE command to set a malformed file attribute under the label `security.NTACL`. SMB2_QUERY_INFO_HE command in testcase trigger the following overflow. [ 4712.003781] ================================================================== [ 4712.003790] BUG: KASAN: slab-out-of-bounds in build_sec_desc+0x842/0x1dd0 [ksmbd] [ 4712.003807] Write of size 1060 at addr ffff88801e34c068 by task kworker/0:0/4190 [ 4712.003813] CPU: 0 PID: 4190 Comm: kworker/0:0 Not tainted 5.19.0-rc5 #1 [ 4712.003850] Workqueue: ksmbd-io handle_ksmbd_work [ksmbd] [ 4712.003867] Call Trace: [ 4712.003870] [ 4712.003873] dump_stack_lvl+0x49/0x5f [ 4712.003935] print_report.cold+0x5e/0x5cf [ 4712.003972] ? ksmbd_vfs_get_sd_xattr+0x16d/0x500 [ksmbd] [ 4712.003984] ? cmp_map_id+0x200/0x200 [ 4712.003988] ? build_sec_desc+0x842/0x1dd0 [ksmbd] [ 4712.004000] kasan_report+0xaa/0x120 [ 4712.004045] ? build_sec_desc+0x842/0x1dd0 [ksmbd] [ 4712.004056] kasan_check_range+0x100/0x1e0 [ 4712.004060] memcpy+0x3c/0x60 [ 4712.004064] build_sec_desc+0x842/0x1dd0 [ksmbd] [ 4712.004076] ? parse_sec_desc+0x580/0x580 [ksmbd] [ 4712.004088] ? ksmbd_acls_fattr+0x281/0x410 [ksmbd] [ 4712.004099] smb2_query_info+0xa8f/0x6110 [ksmbd] [ 4712.004111] ? psi_group_change+0x856/0xd70 [ 4712.004148] ? update_load_avg+0x1c3/0x1af0 [ 4712.004152] ? asym_cpu_capacity_scan+0x5d0/0x5d0 [ 4712.004157] ? xas_load+0x23/0x300 [ 4712.004162] ? smb2_query_dir+0x1530/0x1530 [ksmbd] [ 4712.004173] ? _raw_spin_lock_bh+0xe0/0xe0 [ 4712.004179] handle_ksmbd_work+0x30e/0x1020 [ksmbd] [ 4712.004192] process_one_work+0x778/0x11c0 [ 4712.004227] ? _raw_spin_lock_irq+0x8e/0xe0 [ 4712.004231] worker_thread+0x544/0x1180 [ 4712.004234] ? __cpuidle_text_end+0x4/0x4 [ 4712.004239] kthread+0x282/0x320 [ 4712.004243] ? process_one_work+0x11c0/0x11c0 [ 4712.004246] ? kthread_complete_and_exit+0x30/0x30 [ 4712.004282] ret_from_fork+0x1f/0x30 This patch add the buffer validation for security descriptor that is stored by malformed SMB2_SET_INFO_HE command. and allocate large response buffer about SMB2_O_INFO_SECURITY file info class. Fixes: e2f34481b24d ("cifsd: add server-side procedures for SMB3") Cc: stable@vger.kernel.org Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-17771 Reviewed-by: Hyunchul Lee Signed-off-by: Namjae Jeon Signed-off-by: Steve French Signed-off-by: Greg Kroah-Hartman --- fs/ksmbd/smb2pdu.c | 37 ++++++++----- fs/ksmbd/smbacl.c | 130 ++++++++++++++++++++++++++++++--------------- fs/ksmbd/smbacl.h | 2 +- fs/ksmbd/vfs.c | 5 ++ 4 files changed, 118 insertions(+), 56 deletions(-) diff --git a/fs/ksmbd/smb2pdu.c b/fs/ksmbd/smb2pdu.c index 06552a67a810..28b5d20c8766 100644 --- a/fs/ksmbd/smb2pdu.c +++ b/fs/ksmbd/smb2pdu.c @@ -541,9 +541,10 @@ int smb2_allocate_rsp_buf(struct ksmbd_work *work) struct smb2_query_info_req *req; req = work->request_buf; - if (req->InfoType == SMB2_O_INFO_FILE && - (req->FileInfoClass == FILE_FULL_EA_INFORMATION || - req->FileInfoClass == FILE_ALL_INFORMATION)) + if ((req->InfoType == SMB2_O_INFO_FILE && + (req->FileInfoClass == FILE_FULL_EA_INFORMATION || + req->FileInfoClass == FILE_ALL_INFORMATION)) || + req->InfoType == SMB2_O_INFO_SECURITY) sz = large_sz; } @@ -2981,7 +2982,7 @@ int smb2_open(struct ksmbd_work *work) goto err_out; rc = build_sec_desc(user_ns, - pntsd, NULL, + pntsd, NULL, 0, OWNER_SECINFO | GROUP_SECINFO | DACL_SECINFO, @@ -3824,6 +3825,15 @@ static int verify_info_level(int info_level) return 0; } +static int smb2_resp_buf_len(struct ksmbd_work *work, unsigned short hdr2_len) +{ + int free_len; + + free_len = (int)(work->response_sz - + (get_rfc1002_len(work->response_buf) + 4)) - hdr2_len; + return free_len; +} + static int smb2_calc_max_out_buf_len(struct ksmbd_work *work, unsigned short hdr2_len, unsigned int out_buf_len) @@ -3833,9 +3843,7 @@ static int smb2_calc_max_out_buf_len(struct ksmbd_work *work, if (out_buf_len > work->conn->vals->max_trans_size) return -EINVAL; - free_len = (int)(work->response_sz - - (get_rfc1002_len(work->response_buf) + 4)) - - hdr2_len; + free_len = smb2_resp_buf_len(work, hdr2_len); if (free_len < 0) return -EINVAL; @@ -5087,10 +5095,10 @@ static int smb2_get_info_sec(struct ksmbd_work *work, struct smb_ntsd *pntsd = (struct smb_ntsd *)rsp->Buffer, *ppntsd = NULL; struct smb_fattr fattr = {{0}}; struct inode *inode; - __u32 secdesclen; + __u32 secdesclen = 0; unsigned int id = KSMBD_NO_FID, pid = KSMBD_NO_FID; int addition_info = le32_to_cpu(req->AdditionalInformation); - int rc; + int rc = 0, ppntsd_size = 0; if (addition_info & ~(OWNER_SECINFO | GROUP_SECINFO | DACL_SECINFO | PROTECTED_DACL_SECINFO | @@ -5136,11 +5144,14 @@ static int smb2_get_info_sec(struct ksmbd_work *work, if (test_share_config_flag(work->tcon->share_conf, KSMBD_SHARE_FLAG_ACL_XATTR)) - ksmbd_vfs_get_sd_xattr(work->conn, user_ns, - fp->filp->f_path.dentry, &ppntsd); + ppntsd_size = ksmbd_vfs_get_sd_xattr(work->conn, user_ns, + fp->filp->f_path.dentry, + &ppntsd); - rc = build_sec_desc(user_ns, pntsd, ppntsd, addition_info, - &secdesclen, &fattr); + /* Check if sd buffer size exceeds response buffer size */ + if (smb2_resp_buf_len(work, 8) > ppntsd_size) + rc = build_sec_desc(user_ns, pntsd, ppntsd, ppntsd_size, + addition_info, &secdesclen, &fattr); posix_acl_release(fattr.cf_acls); posix_acl_release(fattr.cf_dacls); kfree(ppntsd); diff --git a/fs/ksmbd/smbacl.c b/fs/ksmbd/smbacl.c index 38f23bf981ac..3781bca2c8fc 100644 --- a/fs/ksmbd/smbacl.c +++ b/fs/ksmbd/smbacl.c @@ -690,6 +690,7 @@ posix_default_acl: static void set_ntacl_dacl(struct user_namespace *user_ns, struct smb_acl *pndacl, struct smb_acl *nt_dacl, + unsigned int aces_size, const struct smb_sid *pownersid, const struct smb_sid *pgrpsid, struct smb_fattr *fattr) @@ -703,9 +704,19 @@ static void set_ntacl_dacl(struct user_namespace *user_ns, if (nt_num_aces) { ntace = (struct smb_ace *)((char *)nt_dacl + sizeof(struct smb_acl)); for (i = 0; i < nt_num_aces; i++) { - memcpy((char *)pndace + size, ntace, le16_to_cpu(ntace->size)); - size += le16_to_cpu(ntace->size); - ntace = (struct smb_ace *)((char *)ntace + le16_to_cpu(ntace->size)); + unsigned short nt_ace_size; + + if (offsetof(struct smb_ace, access_req) > aces_size) + break; + + nt_ace_size = le16_to_cpu(ntace->size); + if (nt_ace_size > aces_size) + break; + + memcpy((char *)pndace + size, ntace, nt_ace_size); + size += nt_ace_size; + aces_size -= nt_ace_size; + ntace = (struct smb_ace *)((char *)ntace + nt_ace_size); num_aces++; } } @@ -878,7 +889,7 @@ int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd, /* Convert permission bits from mode to equivalent CIFS ACL */ int build_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd, struct smb_ntsd *ppntsd, - int addition_info, __u32 *secdesclen, + int ppntsd_size, int addition_info, __u32 *secdesclen, struct smb_fattr *fattr) { int rc = 0; @@ -938,15 +949,25 @@ int build_sec_desc(struct user_namespace *user_ns, if (!ppntsd) { set_mode_dacl(user_ns, dacl_ptr, fattr); - } else if (!ppntsd->dacloffset) { - goto out; } else { struct smb_acl *ppdacl_ptr; + unsigned int dacl_offset = le32_to_cpu(ppntsd->dacloffset); + int ppdacl_size, ntacl_size = ppntsd_size - dacl_offset; + + if (!dacl_offset || + (dacl_offset + sizeof(struct smb_acl) > ppntsd_size)) + goto out; + + ppdacl_ptr = (struct smb_acl *)((char *)ppntsd + dacl_offset); + ppdacl_size = le16_to_cpu(ppdacl_ptr->size); + if (ppdacl_size > ntacl_size || + ppdacl_size < sizeof(struct smb_acl)) + goto out; - ppdacl_ptr = (struct smb_acl *)((char *)ppntsd + - le32_to_cpu(ppntsd->dacloffset)); set_ntacl_dacl(user_ns, dacl_ptr, ppdacl_ptr, - nowner_sid_ptr, ngroup_sid_ptr, fattr); + ntacl_size - sizeof(struct smb_acl), + nowner_sid_ptr, ngroup_sid_ptr, + fattr); } pntsd->dacloffset = cpu_to_le32(offset); offset += le16_to_cpu(dacl_ptr->size); @@ -980,24 +1001,31 @@ int smb_inherit_dacl(struct ksmbd_conn *conn, struct smb_sid owner_sid, group_sid; struct dentry *parent = path->dentry->d_parent; struct user_namespace *user_ns = mnt_user_ns(path->mnt); - int inherited_flags = 0, flags = 0, i, ace_cnt = 0, nt_size = 0; - int rc = 0, num_aces, dacloffset, pntsd_type, acl_len; + int inherited_flags = 0, flags = 0, i, ace_cnt = 0, nt_size = 0, pdacl_size; + int rc = 0, num_aces, dacloffset, pntsd_type, pntsd_size, acl_len, aces_size; char *aces_base; bool is_dir = S_ISDIR(d_inode(path->dentry)->i_mode); - acl_len = ksmbd_vfs_get_sd_xattr(conn, user_ns, - parent, &parent_pntsd); - if (acl_len <= 0) + pntsd_size = ksmbd_vfs_get_sd_xattr(conn, user_ns, + parent, &parent_pntsd); + if (pntsd_size <= 0) return -ENOENT; dacloffset = le32_to_cpu(parent_pntsd->dacloffset); - if (!dacloffset) { + if (!dacloffset || (dacloffset + sizeof(struct smb_acl) > pntsd_size)) { rc = -EINVAL; goto free_parent_pntsd; } parent_pdacl = (struct smb_acl *)((char *)parent_pntsd + dacloffset); + acl_len = pntsd_size - dacloffset; num_aces = le32_to_cpu(parent_pdacl->num_aces); pntsd_type = le16_to_cpu(parent_pntsd->type); + pdacl_size = le16_to_cpu(parent_pdacl->size); + + if (pdacl_size > acl_len || pdacl_size < sizeof(struct smb_acl)) { + rc = -EINVAL; + goto free_parent_pntsd; + } aces_base = kmalloc(sizeof(struct smb_ace) * num_aces * 2, GFP_KERNEL); if (!aces_base) { @@ -1008,11 +1036,23 @@ int smb_inherit_dacl(struct ksmbd_conn *conn, aces = (struct smb_ace *)aces_base; parent_aces = (struct smb_ace *)((char *)parent_pdacl + sizeof(struct smb_acl)); + aces_size = acl_len - sizeof(struct smb_acl); if (pntsd_type & DACL_AUTO_INHERITED) inherited_flags = INHERITED_ACE; for (i = 0; i < num_aces; i++) { + int pace_size; + + if (offsetof(struct smb_ace, access_req) > aces_size) + break; + + pace_size = le16_to_cpu(parent_aces->size); + if (pace_size > aces_size) + break; + + aces_size -= pace_size; + flags = parent_aces->flags; if (!smb_inherit_flags(flags, is_dir)) goto pass; @@ -1057,8 +1097,7 @@ int smb_inherit_dacl(struct ksmbd_conn *conn, aces = (struct smb_ace *)((char *)aces + le16_to_cpu(aces->size)); ace_cnt++; pass: - parent_aces = - (struct smb_ace *)((char *)parent_aces + le16_to_cpu(parent_aces->size)); + parent_aces = (struct smb_ace *)((char *)parent_aces + pace_size); } if (nt_size > 0) { @@ -1153,7 +1192,7 @@ int smb_check_perm_dacl(struct ksmbd_conn *conn, struct path *path, struct smb_ntsd *pntsd = NULL; struct smb_acl *pdacl; struct posix_acl *posix_acls; - int rc = 0, acl_size; + int rc = 0, pntsd_size, acl_size, aces_size, pdacl_size, dacl_offset; struct smb_sid sid; int granted = le32_to_cpu(*pdaccess & ~FILE_MAXIMAL_ACCESS_LE); struct smb_ace *ace; @@ -1162,37 +1201,33 @@ int smb_check_perm_dacl(struct ksmbd_conn *conn, struct path *path, struct smb_ace *others_ace = NULL; struct posix_acl_entry *pa_entry; unsigned int sid_type = SIDOWNER; - char *end_of_acl; + unsigned short ace_size; ksmbd_debug(SMB, "check permission using windows acl\n"); - acl_size = ksmbd_vfs_get_sd_xattr(conn, user_ns, - path->dentry, &pntsd); - if (acl_size <= 0 || !pntsd || !pntsd->dacloffset) { - kfree(pntsd); - return 0; - } + pntsd_size = ksmbd_vfs_get_sd_xattr(conn, user_ns, + path->dentry, &pntsd); + if (pntsd_size <= 0 || !pntsd) + goto err_out; + + dacl_offset = le32_to_cpu(pntsd->dacloffset); + if (!dacl_offset || + (dacl_offset + sizeof(struct smb_acl) > pntsd_size)) + goto err_out; pdacl = (struct smb_acl *)((char *)pntsd + le32_to_cpu(pntsd->dacloffset)); - end_of_acl = ((char *)pntsd) + acl_size; - if (end_of_acl <= (char *)pdacl) { - kfree(pntsd); - return 0; - } + acl_size = pntsd_size - dacl_offset; + pdacl_size = le16_to_cpu(pdacl->size); - if (end_of_acl < (char *)pdacl + le16_to_cpu(pdacl->size) || - le16_to_cpu(pdacl->size) < sizeof(struct smb_acl)) { - kfree(pntsd); - return 0; - } + if (pdacl_size > acl_size || pdacl_size < sizeof(struct smb_acl)) + goto err_out; if (!pdacl->num_aces) { - if (!(le16_to_cpu(pdacl->size) - sizeof(struct smb_acl)) && + if (!(pdacl_size - sizeof(struct smb_acl)) && *pdaccess & ~(FILE_READ_CONTROL_LE | FILE_WRITE_DAC_LE)) { rc = -EACCES; goto err_out; } - kfree(pntsd); - return 0; + goto err_out; } if (*pdaccess & FILE_MAXIMAL_ACCESS_LE) { @@ -1200,11 +1235,16 @@ int smb_check_perm_dacl(struct ksmbd_conn *conn, struct path *path, DELETE; ace = (struct smb_ace *)((char *)pdacl + sizeof(struct smb_acl)); + aces_size = acl_size - sizeof(struct smb_acl); for (i = 0; i < le32_to_cpu(pdacl->num_aces); i++) { + if (offsetof(struct smb_ace, access_req) > aces_size) + break; + ace_size = le16_to_cpu(ace->size); + if (ace_size > aces_size) + break; + aces_size -= ace_size; granted |= le32_to_cpu(ace->access_req); ace = (struct smb_ace *)((char *)ace + le16_to_cpu(ace->size)); - if (end_of_acl < (char *)ace) - goto err_out; } if (!pdacl->num_aces) @@ -1216,7 +1256,15 @@ int smb_check_perm_dacl(struct ksmbd_conn *conn, struct path *path, id_to_sid(uid, sid_type, &sid); ace = (struct smb_ace *)((char *)pdacl + sizeof(struct smb_acl)); + aces_size = acl_size - sizeof(struct smb_acl); for (i = 0; i < le32_to_cpu(pdacl->num_aces); i++) { + if (offsetof(struct smb_ace, access_req) > aces_size) + break; + ace_size = le16_to_cpu(ace->size); + if (ace_size > aces_size) + break; + aces_size -= ace_size; + if (!compare_sids(&sid, &ace->sid) || !compare_sids(&sid_unix_NFS_mode, &ace->sid)) { found = 1; @@ -1226,8 +1274,6 @@ int smb_check_perm_dacl(struct ksmbd_conn *conn, struct path *path, others_ace = ace; ace = (struct smb_ace *)((char *)ace + le16_to_cpu(ace->size)); - if (end_of_acl < (char *)ace) - goto err_out; } if (*pdaccess & FILE_MAXIMAL_ACCESS_LE && found) { diff --git a/fs/ksmbd/smbacl.h b/fs/ksmbd/smbacl.h index 811af3309429..fcb2c83f2992 100644 --- a/fs/ksmbd/smbacl.h +++ b/fs/ksmbd/smbacl.h @@ -193,7 +193,7 @@ struct posix_acl_state { int parse_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd, int acl_len, struct smb_fattr *fattr); int build_sec_desc(struct user_namespace *user_ns, struct smb_ntsd *pntsd, - struct smb_ntsd *ppntsd, int addition_info, + struct smb_ntsd *ppntsd, int ppntsd_size, int addition_info, __u32 *secdesclen, struct smb_fattr *fattr); int init_acl_state(struct posix_acl_state *state, int cnt); void free_acl_state(struct posix_acl_state *state); diff --git a/fs/ksmbd/vfs.c b/fs/ksmbd/vfs.c index 2139aa042c79..513989b1c8cd 100644 --- a/fs/ksmbd/vfs.c +++ b/fs/ksmbd/vfs.c @@ -1543,6 +1543,11 @@ int ksmbd_vfs_get_sd_xattr(struct ksmbd_conn *conn, } *pntsd = acl.sd_buf; + if (acl.sd_size < sizeof(struct smb_ntsd)) { + pr_err("sd size is invalid\n"); + goto out_free; + } + (*pntsd)->osidoffset = cpu_to_le32(le32_to_cpu((*pntsd)->osidoffset) - NDR_NTSD_OFFSETOF); (*pntsd)->gsidoffset = cpu_to_le32(le32_to_cpu((*pntsd)->gsidoffset) - From f6632763484c6078f65eff3fd0044cc2bc82fd18 Mon Sep 17 00:00:00 2001 From: Thadeu Lima de Souza Cascardo Date: Tue, 16 Aug 2022 05:26:56 -0300 Subject: [PATCH 06/11] Revert "x86/ftrace: Use alternative RET encoding" This reverts commit e54fcb0812faebd147de72bd37ad87cc4951c68c. This temporarily reverts the backport of upstream commit 1f001e9da6bbf482311e45e48f53c2bd2179e59c. It was not correct to copy the ftrace stub as it would contain a relative jump to the return thunk which would not apply to the context where it was being copied to, leading to ftrace support to be broken. Signed-off-by: Thadeu Lima de Souza Cascardo Signed-off-by: Greg Kroah-Hartman --- arch/x86/kernel/ftrace.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c index ceab28277546..5080f578236a 100644 --- a/arch/x86/kernel/ftrace.c +++ b/arch/x86/kernel/ftrace.c @@ -309,7 +309,7 @@ union ftrace_op_code_union { } __attribute__((packed)); }; -#define RET_SIZE (IS_ENABLED(CONFIG_RETPOLINE) ? 5 : 1 + IS_ENABLED(CONFIG_SLS)) +#define RET_SIZE 1 + IS_ENABLED(CONFIG_SLS) static unsigned long create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) @@ -368,10 +368,7 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) /* The trampoline ends with ret(q) */ retq = (unsigned long)ftrace_stub; - if (cpu_feature_enabled(X86_FEATURE_RETHUNK)) - memcpy(ip, text_gen_insn(JMP32_INSN_OPCODE, ip, &__x86_return_thunk), JMP32_INSN_SIZE); - else - ret = copy_from_kernel_nofault(ip, (void *)retq, RET_SIZE); + ret = copy_from_kernel_nofault(ip, (void *)retq, RET_SIZE); if (WARN_ON(ret < 0)) goto fail; From 543138c555185e5054f9095909761f1bca9096ba Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 16 Aug 2022 05:26:57 -0300 Subject: [PATCH 07/11] x86/ibt,ftrace: Make function-graph play nice commit e52fc2cf3f662828cc0d51c4b73bed73ad275fce upstream. Return trampoline must not use indirect branch to return; while this preserves the RSB, it is fundamentally incompatible with IBT. Instead use a retpoline like ROP gadget that defeats IBT while not unbalancing the RSB. And since ftrace_stub is no longer a plain RET, don't use it to copy from. Since RET is a trivial instruction, poke it directly. Signed-off-by: Peter Zijlstra (Intel) Acked-by: Josh Poimboeuf Link: https://lore.kernel.org/r/20220308154318.347296408@infradead.org [cascardo: remove ENDBR] Signed-off-by: Thadeu Lima de Souza Cascardo Signed-off-by: Greg Kroah-Hartman --- arch/x86/kernel/ftrace.c | 9 ++------- arch/x86/kernel/ftrace_64.S | 19 +++++++++++++++---- 2 files changed, 17 insertions(+), 11 deletions(-) diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c index 5080f578236a..8160d1dc6ed3 100644 --- a/arch/x86/kernel/ftrace.c +++ b/arch/x86/kernel/ftrace.c @@ -322,12 +322,12 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) unsigned long offset; unsigned long npages; unsigned long size; - unsigned long retq; unsigned long *ptr; void *trampoline; void *ip; /* 48 8b 15 is movq (%rip), %rdx */ unsigned const char op_ref[] = { 0x48, 0x8b, 0x15 }; + unsigned const char retq[] = { RET_INSN_OPCODE, INT3_INSN_OPCODE }; union ftrace_op_code_union op_ptr; int ret; @@ -365,12 +365,7 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) goto fail; ip = trampoline + size; - - /* The trampoline ends with ret(q) */ - retq = (unsigned long)ftrace_stub; - ret = copy_from_kernel_nofault(ip, (void *)retq, RET_SIZE); - if (WARN_ON(ret < 0)) - goto fail; + memcpy(ip, retq, RET_SIZE); /* No need to test direct calls on created trampolines */ if (ops->flags & FTRACE_OPS_FL_SAVE_REGS) { diff --git a/arch/x86/kernel/ftrace_64.S b/arch/x86/kernel/ftrace_64.S index d6af81d1b788..6cc14a835991 100644 --- a/arch/x86/kernel/ftrace_64.S +++ b/arch/x86/kernel/ftrace_64.S @@ -181,7 +181,6 @@ SYM_INNER_LABEL(ftrace_graph_call, SYM_L_GLOBAL) /* * This is weak to keep gas from relaxing the jumps. - * It is also used to copy the RET for trampolines. */ SYM_INNER_LABEL_ALIGN(ftrace_stub, SYM_L_WEAK) UNWIND_HINT_FUNC @@ -335,7 +334,7 @@ SYM_FUNC_START(ftrace_graph_caller) SYM_FUNC_END(ftrace_graph_caller) SYM_FUNC_START(return_to_handler) - subq $24, %rsp + subq $16, %rsp /* Save the return values */ movq %rax, (%rsp) @@ -347,7 +346,19 @@ SYM_FUNC_START(return_to_handler) movq %rax, %rdi movq 8(%rsp), %rdx movq (%rsp), %rax - addq $24, %rsp - JMP_NOSPEC rdi + + addq $16, %rsp + /* + * Jump back to the old return address. This cannot be JMP_NOSPEC rdi + * since IBT would demand that contain ENDBR, which simply isn't so for + * return addresses. Use a retpoline here to keep the RSB balanced. + */ + ANNOTATE_INTRA_FUNCTION_CALL + call .Ldo_rop + int3 +.Ldo_rop: + mov %rdi, (%rsp) + UNWIND_HINT_FUNC + RET SYM_FUNC_END(return_to_handler) #endif From 3eb602ad6a94a76941f93173131a71ad36fa1324 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 16 Aug 2022 05:26:58 -0300 Subject: [PATCH 08/11] x86/ftrace: Use alternative RET encoding commit 1f001e9da6bbf482311e45e48f53c2bd2179e59c upstream. Use the return thunk in ftrace trampolines, if needed. Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Borislav Petkov Reviewed-by: Josh Poimboeuf Signed-off-by: Borislav Petkov [cascardo: use memcpy(text_gen_insn) as there is no __text_gen_insn] Signed-off-by: Thadeu Lima de Souza Cascardo Signed-off-by: Greg Kroah-Hartman --- arch/x86/kernel/ftrace.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c index 8160d1dc6ed3..b3c9ef01d6c0 100644 --- a/arch/x86/kernel/ftrace.c +++ b/arch/x86/kernel/ftrace.c @@ -309,7 +309,7 @@ union ftrace_op_code_union { } __attribute__((packed)); }; -#define RET_SIZE 1 + IS_ENABLED(CONFIG_SLS) +#define RET_SIZE (IS_ENABLED(CONFIG_RETPOLINE) ? 5 : 1 + IS_ENABLED(CONFIG_SLS)) static unsigned long create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) @@ -365,7 +365,12 @@ create_trampoline(struct ftrace_ops *ops, unsigned int *tramp_size) goto fail; ip = trampoline + size; - memcpy(ip, retq, RET_SIZE); + + /* The trampoline ends with ret(q) */ + if (cpu_feature_enabled(X86_FEATURE_RETHUNK)) + memcpy(ip, text_gen_insn(JMP32_INSN_OPCODE, ip, &__x86_return_thunk), JMP32_INSN_SIZE); + else + memcpy(ip, retq, sizeof(retq)); /* No need to test direct calls on created trampolines */ if (ops->flags & FTRACE_OPS_FL_SAVE_REGS) { From 0d9c713cc30f78f5af731afedb91f04b9b6ae70d Mon Sep 17 00:00:00 2001 From: Qu Wenruo Date: Fri, 19 Aug 2022 16:39:49 +0800 Subject: [PATCH 09/11] btrfs: only write the sectors in the vertical stripe which has data stripes commit bd8f7e627703ca5707833d623efcd43f104c7b3f upstream. If we have only 8K partial write at the beginning of a full RAID56 stripe, we will write the following contents: 0 8K 32K 64K Disk 1 (data): |XX| | | Disk 2 (data): | | | Disk 3 (parity): |XXXXXXXXXXXXXXX|XXXXXXXXXXXXXXX| |X| means the sector will be written back to disk. Note that, although we won't write any sectors from disk 2, but we will write the full 64KiB of parity to disk. This behavior is fine for now, but not for the future (especially for RAID56J, as we waste quite some space to journal the unused parity stripes). So here we will also utilize the btrfs_raid_bio::dbitmap, anytime we queue a higher level bio into an rbio, we will update rbio::dbitmap to indicate which vertical stripes we need to writeback. And at finish_rmw(), we also check dbitmap to see if we need to write any sector in the vertical stripe. So after the patch, above example will only lead to the following writeback pattern: 0 8K 32K 64K Disk 1 (data): |XX| | | Disk 2 (data): | | | Disk 3 (parity): |XX| | | Acked-by: David Sterba Signed-off-by: Qu Wenruo Signed-off-by: David Sterba Signed-off-by: Greg Kroah-Hartman --- fs/btrfs/raid56.c | 55 +++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 51 insertions(+), 4 deletions(-) diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c index 893d93e3c516..13a7bab7ff4e 100644 --- a/fs/btrfs/raid56.c +++ b/fs/btrfs/raid56.c @@ -324,6 +324,9 @@ static void merge_rbio(struct btrfs_raid_bio *dest, { bio_list_merge(&dest->bio_list, &victim->bio_list); dest->bio_list_bytes += victim->bio_list_bytes; + /* Also inherit the bitmaps from @victim. */ + bitmap_or(dest->dbitmap, victim->dbitmap, dest->dbitmap, + dest->stripe_npages); dest->generic_bio_cnt += victim->generic_bio_cnt; bio_list_init(&victim->bio_list); } @@ -865,6 +868,12 @@ static void rbio_orig_end_io(struct btrfs_raid_bio *rbio, blk_status_t err) if (rbio->generic_bio_cnt) btrfs_bio_counter_sub(rbio->fs_info, rbio->generic_bio_cnt); + /* + * Clear the data bitmap, as the rbio may be cached for later usage. + * do this before before unlock_stripe() so there will be no new bio + * for this bio. + */ + bitmap_clear(rbio->dbitmap, 0, rbio->stripe_npages); /* * At this moment, rbio->bio_list is empty, however since rbio does not @@ -1197,6 +1206,9 @@ static noinline void finish_rmw(struct btrfs_raid_bio *rbio) else BUG(); + /* We should have at least one data sector. */ + ASSERT(bitmap_weight(rbio->dbitmap, rbio->stripe_npages)); + /* at this point we either have a full stripe, * or we've read the full stripe from the drive. * recalculate the parity and write the new results. @@ -1268,6 +1280,11 @@ static noinline void finish_rmw(struct btrfs_raid_bio *rbio) for (stripe = 0; stripe < rbio->real_stripes; stripe++) { for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) { struct page *page; + + /* This vertical stripe has no data, skip it. */ + if (!test_bit(pagenr, rbio->dbitmap)) + continue; + if (stripe < rbio->nr_data) { page = page_in_rbio(rbio, stripe, pagenr, 1); if (!page) @@ -1292,6 +1309,11 @@ static noinline void finish_rmw(struct btrfs_raid_bio *rbio) for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) { struct page *page; + + /* This vertical stripe has no data, skip it. */ + if (!test_bit(pagenr, rbio->dbitmap)) + continue; + if (stripe < rbio->nr_data) { page = page_in_rbio(rbio, stripe, pagenr, 1); if (!page) @@ -1715,6 +1737,33 @@ static void btrfs_raid_unplug(struct blk_plug_cb *cb, bool from_schedule) run_plug(plug); } +/* Add the original bio into rbio->bio_list, and update rbio::dbitmap. */ +static void rbio_add_bio(struct btrfs_raid_bio *rbio, struct bio *orig_bio) +{ + const struct btrfs_fs_info *fs_info = rbio->fs_info; + const u64 orig_logical = orig_bio->bi_iter.bi_sector << SECTOR_SHIFT; + const u64 full_stripe_start = rbio->bioc->raid_map[0]; + const u32 orig_len = orig_bio->bi_iter.bi_size; + const u32 sectorsize = fs_info->sectorsize; + u64 cur_logical; + + ASSERT(orig_logical >= full_stripe_start && + orig_logical + orig_len <= full_stripe_start + + rbio->nr_data * rbio->stripe_len); + + bio_list_add(&rbio->bio_list, orig_bio); + rbio->bio_list_bytes += orig_bio->bi_iter.bi_size; + + /* Update the dbitmap. */ + for (cur_logical = orig_logical; cur_logical < orig_logical + orig_len; + cur_logical += sectorsize) { + int bit = ((u32)(cur_logical - full_stripe_start) >> + fs_info->sectorsize_bits) % rbio->stripe_npages; + + set_bit(bit, rbio->dbitmap); + } +} + /* * our main entry point for writes from the rest of the FS. */ @@ -1731,9 +1780,8 @@ int raid56_parity_write(struct btrfs_fs_info *fs_info, struct bio *bio, btrfs_put_bioc(bioc); return PTR_ERR(rbio); } - bio_list_add(&rbio->bio_list, bio); - rbio->bio_list_bytes = bio->bi_iter.bi_size; rbio->operation = BTRFS_RBIO_WRITE; + rbio_add_bio(rbio, bio); btrfs_bio_counter_inc_noblocked(fs_info); rbio->generic_bio_cnt = 1; @@ -2135,8 +2183,7 @@ int raid56_parity_recover(struct btrfs_fs_info *fs_info, struct bio *bio, } rbio->operation = BTRFS_RBIO_READ_REBUILD; - bio_list_add(&rbio->bio_list, bio); - rbio->bio_list_bytes = bio->bi_iter.bi_size; + rbio_add_bio(rbio, bio); rbio->faila = find_logical_bio_stripe(rbio, bio); if (rbio->faila == -1) { From 2a9114b3ec6f9d195e7744e50dc13937a88d48db Mon Sep 17 00:00:00 2001 From: Qu Wenruo Date: Fri, 19 Aug 2022 16:39:50 +0800 Subject: [PATCH 10/11] btrfs: raid56: don't trust any cached sector in __raid56_parity_recover() commit f6065f8edeb25f4a9dfe0b446030ad995a84a088 upstream. [BUG] There is a small workload which will always fail with recent kernel: (A simplified version from btrfs/125 test case) mkfs.btrfs -f -m raid5 -d raid5 -b 1G $dev1 $dev2 $dev3 mount $dev1 $mnt xfs_io -f -c "pwrite -S 0xee 0 1M" $mnt/file1 sync umount $mnt btrfs dev scan -u $dev3 mount -o degraded $dev1 $mnt xfs_io -f -c "pwrite -S 0xff 0 128M" $mnt/file2 umount $mnt btrfs dev scan mount $dev1 $mnt btrfs balance start --full-balance $mnt umount $mnt The failure is always failed to read some tree blocks: BTRFS info (device dm-4): relocating block group 217710592 flags data|raid5 BTRFS error (device dm-4): parent transid verify failed on 38993920 wanted 9 found 7 BTRFS error (device dm-4): parent transid verify failed on 38993920 wanted 9 found 7 ... [CAUSE] With the recently added debug output, we can see all RAID56 operations related to full stripe 38928384: 56.1183: raid56_read_partial: full_stripe=38928384 devid=2 type=DATA1 offset=0 opf=0x0 physical=9502720 len=65536 56.1185: raid56_read_partial: full_stripe=38928384 devid=3 type=DATA2 offset=16384 opf=0x0 physical=9519104 len=16384 56.1185: raid56_read_partial: full_stripe=38928384 devid=3 type=DATA2 offset=49152 opf=0x0 physical=9551872 len=16384 56.1187: raid56_write_stripe: full_stripe=38928384 devid=3 type=DATA2 offset=0 opf=0x1 physical=9502720 len=16384 56.1188: raid56_write_stripe: full_stripe=38928384 devid=3 type=DATA2 offset=32768 opf=0x1 physical=9535488 len=16384 56.1188: raid56_write_stripe: full_stripe=38928384 devid=1 type=PQ1 offset=0 opf=0x1 physical=30474240 len=16384 56.1189: raid56_write_stripe: full_stripe=38928384 devid=1 type=PQ1 offset=32768 opf=0x1 physical=30507008 len=16384 56.1218: raid56_write_stripe: full_stripe=38928384 devid=3 type=DATA2 offset=49152 opf=0x1 physical=9551872 len=16384 56.1219: raid56_write_stripe: full_stripe=38928384 devid=1 type=PQ1 offset=49152 opf=0x1 physical=30523392 len=16384 56.2721: raid56_parity_recover: full stripe=38928384 eb=39010304 mirror=2 56.2723: raid56_parity_recover: full stripe=38928384 eb=39010304 mirror=2 56.2724: raid56_parity_recover: full stripe=38928384 eb=39010304 mirror=2 Before we enter raid56_parity_recover(), we have triggered some metadata write for the full stripe 38928384, this leads to us to read all the sectors from disk. Furthermore, btrfs raid56 write will cache its calculated P/Q sectors to avoid unnecessary read. This means, for that full stripe, after any partial write, we will have stale data, along with P/Q calculated using that stale data. Thankfully due to patch "btrfs: only write the sectors in the vertical stripe which has data stripes" we haven't submitted all the corrupted P/Q to disk. When we really need to recover certain range, aka in raid56_parity_recover(), we will use the cached rbio, along with its cached sectors (the full stripe is all cached). This explains why we have no event raid56_scrub_read_recover() triggered. Since we have the cached P/Q which is calculated using the stale data, the recovered one will just be stale. In our particular test case, it will always return the same incorrect metadata, thus causing the same error message "parent transid verify failed on 39010304 wanted 9 found 7" again and again. [BTRFS DESTRUCTIVE RMW PROBLEM] Test case btrfs/125 (and above workload) always has its trouble with the destructive read-modify-write (RMW) cycle: 0 32K 64K Data1: | Good | Good | Data2: | Bad | Bad | Parity: | Good | Good | In above case, if we trigger any write into Data1, we will use the bad data in Data2 to re-generate parity, killing the only chance to recovery Data2, thus Data2 is lost forever. This destructive RMW cycle is not specific to btrfs RAID56, but there are some btrfs specific behaviors making the case even worse: - Btrfs will cache sectors for unrelated vertical stripes. In above example, if we're only writing into 0~32K range, btrfs will still read data range (32K ~ 64K) of Data1, and (64K~128K) of Data2. This behavior is to cache sectors for later update. Incidentally commit d4e28d9b5f04 ("btrfs: raid56: make steal_rbio() subpage compatible") has a bug which makes RAID56 to never trust the cached sectors, thus slightly improve the situation for recovery. Unfortunately, follow up fix "btrfs: update stripe_sectors::uptodate in steal_rbio" will revert the behavior back to the old one. - Btrfs raid56 partial write will update all P/Q sectors and cache them This means, even if data at (64K ~ 96K) of Data2 is free space, and only (96K ~ 128K) of Data2 is really stale data. And we write into that (96K ~ 128K), we will update all the parity sectors for the full stripe. This unnecessary behavior will completely kill the chance of recovery. Thankfully, an unrelated optimization "btrfs: only write the sectors in the vertical stripe which has data stripes" will prevent submitting the write bio for untouched vertical sectors. That optimization will keep the on-disk P/Q untouched for a chance for later recovery. [FIX] Although we have no good way to completely fix the destructive RMW (unless we go full scrub for each partial write), we can still limit the damage. With patch "btrfs: only write the sectors in the vertical stripe which has data stripes" now we won't really submit the P/Q of unrelated vertical stripes, so the on-disk P/Q should still be fine. Now we really need to do is just drop all the cached sectors when doing recovery. By this, we have a chance to read the original P/Q from disk, and have a chance to recover the stale data, while still keep the cache to speed up regular write path. In fact, just dropping all the cache for recovery path is good enough to allow the test case btrfs/125 along with the small script to pass reliably. The lack of metadata write after the degraded mount, and forced metadata COW is saving us this time. So this patch will fix the behavior by not trust any cache in __raid56_parity_recover(), to solve the problem while still keep the cache useful. But please note that this test pass DOES NOT mean we have solved the destructive RMW problem, we just do better damage control a little better. Related patches: - btrfs: only write the sectors in the vertical stripe - d4e28d9b5f04 ("btrfs: raid56: make steal_rbio() subpage compatible") - btrfs: update stripe_sectors::uptodate in steal_rbio Acked-by: David Sterba Signed-off-by: Qu Wenruo Signed-off-by: David Sterba Signed-off-by: Greg Kroah-Hartman --- fs/btrfs/raid56.c | 19 ++++++------------- 1 file changed, 6 insertions(+), 13 deletions(-) diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c index 13a7bab7ff4e..3157a26ddf7e 100644 --- a/fs/btrfs/raid56.c +++ b/fs/btrfs/raid56.c @@ -2085,9 +2085,12 @@ static int __raid56_parity_recover(struct btrfs_raid_bio *rbio) atomic_set(&rbio->error, 0); /* - * read everything that hasn't failed. Thanks to the - * stripe cache, it is possible that some or all of these - * pages are going to be uptodate. + * Read everything that hasn't failed. However this time we will + * not trust any cached sector. + * As we may read out some stale data but higher layer is not reading + * that stale part. + * + * So here we always re-read everything in recovery path. */ for (stripe = 0; stripe < rbio->real_stripes; stripe++) { if (rbio->faila == stripe || rbio->failb == stripe) { @@ -2096,16 +2099,6 @@ static int __raid56_parity_recover(struct btrfs_raid_bio *rbio) } for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) { - struct page *p; - - /* - * the rmw code may have already read this - * page in - */ - p = rbio_stripe_page(rbio, stripe, pagenr); - if (PageUptodate(p)) - continue; - ret = rbio_add_io_page(rbio, &bio_list, rbio_stripe_page(rbio, stripe, pagenr), stripe, pagenr, rbio->stripe_len); From a0a7e0b2b8b22901945ea2aef1b65871d718accf Mon Sep 17 00:00:00 2001 From: Greg Kroah-Hartman Date: Sun, 21 Aug 2022 15:17:49 +0200 Subject: [PATCH 11/11] Linux 5.15.62 Link: https://lore.kernel.org/r/20220819153711.658766010@linuxfoundation.org Tested-by: Shuah Khan Tested-by: Bagas Sanjaya Tested-by: Sudip Mukherjee Tested-by: Ron Economos Link: https://lore.kernel.org/r/20220820182309.607584465@linuxfoundation.org Tested-by: Guenter Roeck Signed-off-by: Greg Kroah-Hartman --- Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Makefile b/Makefile index 0802acf352d2..5b4f8f8851bf 100644 --- a/Makefile +++ b/Makefile @@ -1,7 +1,7 @@ # SPDX-License-Identifier: GPL-2.0 VERSION = 5 PATCHLEVEL = 15 -SUBLEVEL = 61 +SUBLEVEL = 62 EXTRAVERSION = NAME = Trick or Treat