From 56d5f3eba3f5de0efdd556de4ef381e109b973a9 Mon Sep 17 00:00:00 2001 From: Christian Brauner Date: Tue, 11 Feb 2025 18:15:59 +0100 Subject: [PATCH 1/5] acct: perform last write from workqueue In [1] it was reported that the acct(2) system call can be used to trigger NULL deref in cases where it is set to write to a file that triggers an internal lookup. This can e.g., happen when pointing acc(2) to /sys/power/resume. At the point the where the write to this file happens the calling task has already exited and called exit_fs(). A lookup will thus trigger a NULL-deref when accessing current->fs. Reorganize the code so that the the final write happens from the workqueue but with the caller's credentials. This preserves the (strange) permission model and has almost no regression risk. This api should stop to exist though. Link: https://lore.kernel.org/r/20250127091811.3183623-1-quzicheng@huawei.com [1] Link: https://lore.kernel.org/r/20250211-work-acct-v1-1-1c16aecab8b3@kernel.org Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reported-by: Zicheng Qu Cc: stable@vger.kernel.org Signed-off-by: Christian Brauner --- kernel/acct.c | 120 +++++++++++++++++++++++++++++--------------------- 1 file changed, 70 insertions(+), 50 deletions(-) diff --git a/kernel/acct.c b/kernel/acct.c index 31222e8cd534..48283efe8a12 100644 --- a/kernel/acct.c +++ b/kernel/acct.c @@ -103,48 +103,50 @@ struct bsd_acct_struct { atomic_long_t count; struct rcu_head rcu; struct mutex lock; - int active; + bool active; + bool check_space; unsigned long needcheck; struct file *file; struct pid_namespace *ns; struct work_struct work; struct completion done; + acct_t ac; }; -static void do_acct_process(struct bsd_acct_struct *acct); +static void fill_ac(struct bsd_acct_struct *acct); +static void acct_write_process(struct bsd_acct_struct *acct); /* * Check the amount of free space and suspend/resume accordingly. */ -static int check_free_space(struct bsd_acct_struct *acct) +static bool check_free_space(struct bsd_acct_struct *acct) { struct kstatfs sbuf; - if (time_is_after_jiffies(acct->needcheck)) - goto out; + if (!acct->check_space) + return acct->active; /* May block */ if (vfs_statfs(&acct->file->f_path, &sbuf)) - goto out; + return acct->active; if (acct->active) { u64 suspend = sbuf.f_blocks * SUSPEND; do_div(suspend, 100); if (sbuf.f_bavail <= suspend) { - acct->active = 0; + acct->active = false; pr_info("Process accounting paused\n"); } } else { u64 resume = sbuf.f_blocks * RESUME; do_div(resume, 100); if (sbuf.f_bavail >= resume) { - acct->active = 1; + acct->active = true; pr_info("Process accounting resumed\n"); } } acct->needcheck = jiffies + ACCT_TIMEOUT*HZ; -out: return acct->active; } @@ -189,7 +191,11 @@ static void acct_pin_kill(struct fs_pin *pin) { struct bsd_acct_struct *acct = to_acct(pin); mutex_lock(&acct->lock); - do_acct_process(acct); + /* + * Fill the accounting struct with the exiting task's info + * before punting to the workqueue. + */ + fill_ac(acct); schedule_work(&acct->work); wait_for_completion(&acct->done); cmpxchg(&acct->ns->bacct, pin, NULL); @@ -202,6 +208,9 @@ static void close_work(struct work_struct *work) { struct bsd_acct_struct *acct = container_of(work, struct bsd_acct_struct, work); struct file *file = acct->file; + + /* We were fired by acct_pin_kill() which holds acct->lock. */ + acct_write_process(acct); if (file->f_op->flush) file->f_op->flush(file, NULL); __fput_sync(file); @@ -430,13 +439,27 @@ static u32 encode_float(u64 value) * do_exit() or when switching to a different output file. */ -static void fill_ac(acct_t *ac) +static void fill_ac(struct bsd_acct_struct *acct) { struct pacct_struct *pacct = ¤t->signal->pacct; + struct file *file = acct->file; + acct_t *ac = &acct->ac; u64 elapsed, run_time; time64_t btime; struct tty_struct *tty; + lockdep_assert_held(&acct->lock); + + if (time_is_after_jiffies(acct->needcheck)) { + acct->check_space = false; + + /* Don't fill in @ac if nothing will be written. */ + if (!acct->active) + return; + } else { + acct->check_space = true; + } + /* * Fill the accounting struct with the needed info as recorded * by the different kernel functions. @@ -484,64 +507,61 @@ static void fill_ac(acct_t *ac) ac->ac_majflt = encode_comp_t(pacct->ac_majflt); ac->ac_exitcode = pacct->ac_exitcode; spin_unlock_irq(¤t->sighand->siglock); -} -/* - * do_acct_process does all actual work. Caller holds the reference to file. - */ -static void do_acct_process(struct bsd_acct_struct *acct) -{ - acct_t ac; - unsigned long flim; - const struct cred *orig_cred; - struct file *file = acct->file; - /* - * Accounting records are not subject to resource limits. - */ - flim = rlimit(RLIMIT_FSIZE); - current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; - /* Perform file operations on behalf of whoever enabled accounting */ - orig_cred = override_creds(file->f_cred); - - /* - * First check to see if there is enough free_space to continue - * the process accounting system. - */ - if (!check_free_space(acct)) - goto out; - - fill_ac(&ac); /* we really need to bite the bullet and change layout */ - ac.ac_uid = from_kuid_munged(file->f_cred->user_ns, orig_cred->uid); - ac.ac_gid = from_kgid_munged(file->f_cred->user_ns, orig_cred->gid); + ac->ac_uid = from_kuid_munged(file->f_cred->user_ns, current_uid()); + ac->ac_gid = from_kgid_munged(file->f_cred->user_ns, current_gid()); #if ACCT_VERSION == 1 || ACCT_VERSION == 2 /* backward-compatible 16 bit fields */ - ac.ac_uid16 = ac.ac_uid; - ac.ac_gid16 = ac.ac_gid; + ac->ac_uid16 = ac->ac_uid; + ac->ac_gid16 = ac->ac_gid; #elif ACCT_VERSION == 3 { struct pid_namespace *ns = acct->ns; - ac.ac_pid = task_tgid_nr_ns(current, ns); + ac->ac_pid = task_tgid_nr_ns(current, ns); rcu_read_lock(); - ac.ac_ppid = task_tgid_nr_ns(rcu_dereference(current->real_parent), - ns); + ac->ac_ppid = task_tgid_nr_ns(rcu_dereference(current->real_parent), ns); rcu_read_unlock(); } #endif +} + +static void acct_write_process(struct bsd_acct_struct *acct) +{ + struct file *file = acct->file; + const struct cred *cred; + acct_t *ac = &acct->ac; + + /* Perform file operations on behalf of whoever enabled accounting */ + cred = override_creds(file->f_cred); + /* - * Get freeze protection. If the fs is frozen, just skip the write - * as we could deadlock the system otherwise. + * First check to see if there is enough free_space to continue + * the process accounting system. Then get freeze protection. If + * the fs is frozen, just skip the write as we could deadlock + * the system otherwise. */ - if (file_start_write_trylock(file)) { + if (check_free_space(acct) && file_start_write_trylock(file)) { /* it's been opened O_APPEND, so position is irrelevant */ loff_t pos = 0; - __kernel_write(file, &ac, sizeof(acct_t), &pos); + __kernel_write(file, ac, sizeof(acct_t), &pos); file_end_write(file); } -out: + + revert_creds(cred); +} + +static void do_acct_process(struct bsd_acct_struct *acct) +{ + unsigned long flim; + + /* Accounting records are not subject to resource limits. */ + flim = rlimit(RLIMIT_FSIZE); + current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; + fill_ac(acct); + acct_write_process(acct); current->signal->rlim[RLIMIT_FSIZE].rlim_cur = flim; - revert_creds(orig_cred); } /** From 890ed45bde808c422c3c27d3285fc45affa0f930 Mon Sep 17 00:00:00 2001 From: Christian Brauner Date: Tue, 11 Feb 2025 18:16:00 +0100 Subject: [PATCH 2/5] acct: block access to kernel internal filesystems There's no point in allowing anything kernel internal nor procfs or sysfs. Link: https://lore.kernel.org/r/20250127091811.3183623-1-quzicheng@huawei.com Link: https://lore.kernel.org/r/20250211-work-acct-v1-2-1c16aecab8b3@kernel.org Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reviewed-by: Amir Goldstein Reported-by: Zicheng Qu Cc: stable@vger.kernel.org Signed-off-by: Christian Brauner --- kernel/acct.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/kernel/acct.c b/kernel/acct.c index 48283efe8a12..6520baa13669 100644 --- a/kernel/acct.c +++ b/kernel/acct.c @@ -243,6 +243,20 @@ static int acct_on(struct filename *pathname) return -EACCES; } + /* Exclude kernel kernel internal filesystems. */ + if (file_inode(file)->i_sb->s_flags & (SB_NOUSER | SB_KERNMOUNT)) { + kfree(acct); + filp_close(file, NULL); + return -EINVAL; + } + + /* Exclude procfs and sysfs. */ + if (file_inode(file)->i_sb->s_iflags & SB_I_USERNS_VISIBLE) { + kfree(acct); + filp_close(file, NULL); + return -EINVAL; + } + if (!(file->f_mode & FMODE_CAN_WRITE)) { kfree(acct); filp_close(file, NULL); From 1d0013962d220b166d9f7c9fe2746f1542e459a3 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 12 Feb 2025 22:23:59 +0000 Subject: [PATCH 3/5] netfs: Fix a number of read-retry hangs Fix a number of hangs in the netfslib read-retry code, including: (1) netfs_reissue_read() doubles up the getting of references on subrequests, thereby leaking the subrequest and causing inode eviction to wait indefinitely. This can lead to the kernel reporting a hang in the filesystem's evict_inode(). Fix this by removing the get from netfs_reissue_read() and adding one to netfs_retry_read_subrequests() to deal with the one place that didn't double up. (2) The loop in netfs_retry_read_subrequests() that retries a sequence of failed subrequests doesn't record whether or not it retried the one that the "subreq" pointer points to when it leaves the loop. It may not if renegotiation/repreparation of the subrequests means that fewer subrequests are needed to span the cumulative range of the sequence. Because it doesn't record this, the piece of code that discards now-superfluous subrequests doesn't know whether it should discard the one "subreq" points to - and so it doesn't. Fix this by noting whether the last subreq it examines is superfluous and if it is, then getting rid of it and all subsequent subrequests. If that one one wasn't superfluous, then we would have tried to go round the previous loop again and so there can be no further unretried subrequests in the sequence. (3) netfs_retry_read_subrequests() gets yet an extra ref on any additional subrequests it has to get because it ran out of ones it could reuse to to renegotiation/repreparation shrinking the subrequests. Fix this by removing that extra ref. (4) In netfs_retry_reads(), it was using wait_on_bit() to wait for NETFS_SREQ_IN_PROGRESS to be cleared on all subrequests in the sequence - but netfs_read_subreq_terminated() is now using a wait queue on the request instead and so this wait will never finish. Fix this by waiting on the wait queue instead. To make this work, a new flag, NETFS_RREQ_RETRYING, is now set around the wait loop to tell the wake-up code to wake up the wait queue rather than requeuing the request's work item. Note that this flag replaces the NETFS_RREQ_NEED_RETRY flag which is no longer used. (5) Whilst not strictly anything to do with the hang, netfs_retry_read_subrequests() was also doubly incrementing the subreq_counter and re-setting the debug index, leaving a gap in the trace. This is also fixed. One of these hangs was observed with 9p and with cifs. Others were forced by manual code injection into fs/afs/file.c. Firstly, afs_prepare_read() was created to provide an changing pattern of maximum subrequest sizes: static int afs_prepare_read(struct netfs_io_subrequest *subreq) { struct netfs_io_request *rreq = subreq->rreq; if (!S_ISREG(subreq->rreq->inode->i_mode)) return 0; if (subreq->retry_count < 20) rreq->io_streams[0].sreq_max_len = umax(200, 2222 - subreq->retry_count * 40); else rreq->io_streams[0].sreq_max_len = 3333; return 0; } and pointed to by afs_req_ops. Then the following: struct netfs_io_subrequest *subreq = op->fetch.subreq; if (subreq->error == 0 && S_ISREG(subreq->rreq->inode->i_mode) && subreq->retry_count < 20) { subreq->transferred = subreq->already_done; __clear_bit(NETFS_SREQ_HIT_EOF, &subreq->flags); __set_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags); afs_fetch_data_notify(op); return; } was inserted into afs_fetch_data_success() at the beginning and struct netfs_io_subrequest given an extra field, "already_done" that was set to the value in "subreq->transferred" by netfs_reissue_read(). When reading a 4K file, the subrequests would get gradually smaller, a new subrequest would be allocated around the 3rd retry and then eventually be rendered superfluous when the 20th retry was hit and the limit on the first subrequest was eased. Fixes: e2d46f2ec332 ("netfs: Change the read result collector to only use one work item") Signed-off-by: David Howells Link: https://lore.kernel.org/r/20250212222402.3618494-2-dhowells@redhat.com Tested-by: Marc Dionne Tested-by: Steve French cc: Ihor Solodrai cc: Eric Van Hensbergen cc: Latchesar Ionkov cc: Dominique Martinet cc: Christian Schoenebeck cc: Paulo Alcantara cc: Jeff Layton cc: v9fs@lists.linux.dev cc: linux-cifs@vger.kernel.org cc: netfs@lists.linux.dev cc: linux-fsdevel@vger.kernel.org Signed-off-by: Christian Brauner --- fs/netfs/read_collect.c | 6 ++++-- fs/netfs/read_retry.c | 40 +++++++++++++++++++++++++++--------- include/linux/netfs.h | 2 +- include/trace/events/netfs.h | 4 +++- 4 files changed, 38 insertions(+), 14 deletions(-) diff --git a/fs/netfs/read_collect.c b/fs/netfs/read_collect.c index f65affa5a9e4..636cc5a98ef5 100644 --- a/fs/netfs/read_collect.c +++ b/fs/netfs/read_collect.c @@ -470,7 +470,8 @@ void netfs_read_collection_worker(struct work_struct *work) */ void netfs_wake_read_collector(struct netfs_io_request *rreq) { - if (test_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &rreq->flags)) { + if (test_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &rreq->flags) && + !test_bit(NETFS_RREQ_RETRYING, &rreq->flags)) { if (!work_pending(&rreq->work)) { netfs_get_request(rreq, netfs_rreq_trace_get_work); if (!queue_work(system_unbound_wq, &rreq->work)) @@ -586,7 +587,8 @@ void netfs_read_subreq_terminated(struct netfs_io_subrequest *subreq) smp_mb__after_atomic(); /* Clear IN_PROGRESS before task state */ /* If we are at the head of the queue, wake up the collector. */ - if (list_is_first(&subreq->rreq_link, &stream->subrequests)) + if (list_is_first(&subreq->rreq_link, &stream->subrequests) || + test_bit(NETFS_RREQ_RETRYING, &rreq->flags)) netfs_wake_read_collector(rreq); netfs_put_subrequest(subreq, true, netfs_sreq_trace_put_terminated); diff --git a/fs/netfs/read_retry.c b/fs/netfs/read_retry.c index 2290af0d51ac..8316c4533a51 100644 --- a/fs/netfs/read_retry.c +++ b/fs/netfs/read_retry.c @@ -14,7 +14,6 @@ static void netfs_reissue_read(struct netfs_io_request *rreq, { __clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags); __set_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags); - netfs_get_subrequest(subreq, netfs_sreq_trace_get_resubmit); subreq->rreq->netfs_ops->issue_read(subreq); } @@ -48,6 +47,7 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq) __clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags); subreq->retry_count++; netfs_reset_iter(subreq); + netfs_get_subrequest(subreq, netfs_sreq_trace_get_resubmit); netfs_reissue_read(rreq, subreq); } } @@ -75,7 +75,7 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq) struct iov_iter source; unsigned long long start, len; size_t part; - bool boundary = false; + bool boundary = false, subreq_superfluous = false; /* Go through the subreqs and find the next span of contiguous * buffer that we then rejig (cifs, for example, needs the @@ -116,8 +116,10 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq) /* Work through the sublist. */ subreq = from; list_for_each_entry_from(subreq, &stream->subrequests, rreq_link) { - if (!len) + if (!len) { + subreq_superfluous = true; break; + } subreq->source = NETFS_DOWNLOAD_FROM_SERVER; subreq->start = start - subreq->transferred; subreq->len = len + subreq->transferred; @@ -154,19 +156,21 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq) netfs_get_subrequest(subreq, netfs_sreq_trace_get_resubmit); netfs_reissue_read(rreq, subreq); - if (subreq == to) + if (subreq == to) { + subreq_superfluous = false; break; + } } /* If we managed to use fewer subreqs, we can discard the * excess; if we used the same number, then we're done. */ if (!len) { - if (subreq == to) + if (!subreq_superfluous) continue; list_for_each_entry_safe_from(subreq, tmp, &stream->subrequests, rreq_link) { - trace_netfs_sreq(subreq, netfs_sreq_trace_discard); + trace_netfs_sreq(subreq, netfs_sreq_trace_superfluous); list_del(&subreq->rreq_link); netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_done); if (subreq == to) @@ -187,14 +191,12 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq) subreq->source = NETFS_DOWNLOAD_FROM_SERVER; subreq->start = start; subreq->len = len; - subreq->debug_index = atomic_inc_return(&rreq->subreq_counter); subreq->stream_nr = stream->stream_nr; subreq->retry_count = 1; trace_netfs_sreq_ref(rreq->debug_id, subreq->debug_index, refcount_read(&subreq->ref), netfs_sreq_trace_new); - netfs_get_subrequest(subreq, netfs_sreq_trace_get_resubmit); list_add(&subreq->rreq_link, &to->rreq_link); to = list_next_entry(to, rreq_link); @@ -256,14 +258,32 @@ void netfs_retry_reads(struct netfs_io_request *rreq) { struct netfs_io_subrequest *subreq; struct netfs_io_stream *stream = &rreq->io_streams[0]; + DEFINE_WAIT(myself); + + set_bit(NETFS_RREQ_RETRYING, &rreq->flags); /* Wait for all outstanding I/O to quiesce before performing retries as * we may need to renegotiate the I/O sizes. */ list_for_each_entry(subreq, &stream->subrequests, rreq_link) { - wait_on_bit(&subreq->flags, NETFS_SREQ_IN_PROGRESS, - TASK_UNINTERRUPTIBLE); + if (!test_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags)) + continue; + + trace_netfs_rreq(rreq, netfs_rreq_trace_wait_queue); + for (;;) { + prepare_to_wait(&rreq->waitq, &myself, TASK_UNINTERRUPTIBLE); + + if (!test_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags)) + break; + + trace_netfs_sreq(subreq, netfs_sreq_trace_wait_for); + schedule(); + trace_netfs_rreq(rreq, netfs_rreq_trace_woke_queue); + } + + finish_wait(&rreq->waitq, &myself); } + clear_bit(NETFS_RREQ_RETRYING, &rreq->flags); trace_netfs_rreq(rreq, netfs_rreq_trace_resubmit); netfs_retry_read_subrequests(rreq); diff --git a/include/linux/netfs.h b/include/linux/netfs.h index 071d05d81d38..c86a11cfc4a3 100644 --- a/include/linux/netfs.h +++ b/include/linux/netfs.h @@ -278,7 +278,7 @@ struct netfs_io_request { #define NETFS_RREQ_PAUSE 11 /* Pause subrequest generation */ #define NETFS_RREQ_USE_IO_ITER 12 /* Use ->io_iter rather than ->i_pages */ #define NETFS_RREQ_ALL_QUEUED 13 /* All subreqs are now queued */ -#define NETFS_RREQ_NEED_RETRY 14 /* Need to try retrying */ +#define NETFS_RREQ_RETRYING 14 /* Set if we're in the retry path */ #define NETFS_RREQ_USE_PGPRIV2 31 /* [DEPRECATED] Use PG_private_2 to mark * write to cache on read */ const struct netfs_request_ops *netfs_ops; diff --git a/include/trace/events/netfs.h b/include/trace/events/netfs.h index 6e699cadcb29..f880835f7695 100644 --- a/include/trace/events/netfs.h +++ b/include/trace/events/netfs.h @@ -99,7 +99,7 @@ EM(netfs_sreq_trace_limited, "LIMIT") \ EM(netfs_sreq_trace_need_clear, "N-CLR") \ EM(netfs_sreq_trace_partial_read, "PARTR") \ - EM(netfs_sreq_trace_need_retry, "NRTRY") \ + EM(netfs_sreq_trace_need_retry, "ND-RT") \ EM(netfs_sreq_trace_prepare, "PREP ") \ EM(netfs_sreq_trace_prep_failed, "PRPFL") \ EM(netfs_sreq_trace_progress, "PRGRS") \ @@ -108,7 +108,9 @@ EM(netfs_sreq_trace_short, "SHORT") \ EM(netfs_sreq_trace_split, "SPLIT") \ EM(netfs_sreq_trace_submit, "SUBMT") \ + EM(netfs_sreq_trace_superfluous, "SPRFL") \ EM(netfs_sreq_trace_terminated, "TERM ") \ + EM(netfs_sreq_trace_wait_for, "_WAIT") \ EM(netfs_sreq_trace_write, "WRITE") \ EM(netfs_sreq_trace_write_skip, "SKIP ") \ E_(netfs_sreq_trace_write_term, "WTERM") From d01c495f432ce34df8bfd092e71720a2cf169a90 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 12 Feb 2025 22:24:00 +0000 Subject: [PATCH 4/5] netfs: Add retry stat counters Add stat counters to count the number of request and subrequest retries and display them in /proc/fs/netfs/stats. Signed-off-by: David Howells Link: https://lore.kernel.org/r/20250212222402.3618494-3-dhowells@redhat.com cc: Jeff Layton cc: netfs@lists.linux.dev cc: linux-fsdevel@vger.kernel.org Signed-off-by: Christian Brauner --- fs/netfs/internal.h | 4 ++++ fs/netfs/read_retry.c | 3 +++ fs/netfs/stats.c | 9 +++++++++ fs/netfs/write_issue.c | 1 + fs/netfs/write_retry.c | 2 ++ 5 files changed, 19 insertions(+) diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h index eb76f98c894b..1c4f953c3d68 100644 --- a/fs/netfs/internal.h +++ b/fs/netfs/internal.h @@ -135,6 +135,8 @@ extern atomic_t netfs_n_rh_write_begin; extern atomic_t netfs_n_rh_write_done; extern atomic_t netfs_n_rh_write_failed; extern atomic_t netfs_n_rh_write_zskip; +extern atomic_t netfs_n_rh_retry_read_req; +extern atomic_t netfs_n_rh_retry_read_subreq; extern atomic_t netfs_n_wh_buffered_write; extern atomic_t netfs_n_wh_writethrough; extern atomic_t netfs_n_wh_dio_write; @@ -147,6 +149,8 @@ extern atomic_t netfs_n_wh_upload_failed; extern atomic_t netfs_n_wh_write; extern atomic_t netfs_n_wh_write_done; extern atomic_t netfs_n_wh_write_failed; +extern atomic_t netfs_n_wh_retry_write_req; +extern atomic_t netfs_n_wh_retry_write_subreq; extern atomic_t netfs_n_wb_lock_skip; extern atomic_t netfs_n_wb_lock_wait; extern atomic_t netfs_n_folioq; diff --git a/fs/netfs/read_retry.c b/fs/netfs/read_retry.c index 8316c4533a51..0f294b26e08c 100644 --- a/fs/netfs/read_retry.c +++ b/fs/netfs/read_retry.c @@ -14,6 +14,7 @@ static void netfs_reissue_read(struct netfs_io_request *rreq, { __clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags); __set_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags); + netfs_stat(&netfs_n_rh_retry_read_subreq); subreq->rreq->netfs_ops->issue_read(subreq); } @@ -260,6 +261,8 @@ void netfs_retry_reads(struct netfs_io_request *rreq) struct netfs_io_stream *stream = &rreq->io_streams[0]; DEFINE_WAIT(myself); + netfs_stat(&netfs_n_rh_retry_read_req); + set_bit(NETFS_RREQ_RETRYING, &rreq->flags); /* Wait for all outstanding I/O to quiesce before performing retries as diff --git a/fs/netfs/stats.c b/fs/netfs/stats.c index f1af344266cc..ab6b916addc4 100644 --- a/fs/netfs/stats.c +++ b/fs/netfs/stats.c @@ -29,6 +29,8 @@ atomic_t netfs_n_rh_write_begin; atomic_t netfs_n_rh_write_done; atomic_t netfs_n_rh_write_failed; atomic_t netfs_n_rh_write_zskip; +atomic_t netfs_n_rh_retry_read_req; +atomic_t netfs_n_rh_retry_read_subreq; atomic_t netfs_n_wh_buffered_write; atomic_t netfs_n_wh_writethrough; atomic_t netfs_n_wh_dio_write; @@ -41,6 +43,8 @@ atomic_t netfs_n_wh_upload_failed; atomic_t netfs_n_wh_write; atomic_t netfs_n_wh_write_done; atomic_t netfs_n_wh_write_failed; +atomic_t netfs_n_wh_retry_write_req; +atomic_t netfs_n_wh_retry_write_subreq; atomic_t netfs_n_wb_lock_skip; atomic_t netfs_n_wb_lock_wait; atomic_t netfs_n_folioq; @@ -81,6 +85,11 @@ int netfs_stats_show(struct seq_file *m, void *v) atomic_read(&netfs_n_wh_write), atomic_read(&netfs_n_wh_write_done), atomic_read(&netfs_n_wh_write_failed)); + seq_printf(m, "Retries: rq=%u rs=%u wq=%u ws=%u\n", + atomic_read(&netfs_n_rh_retry_read_req), + atomic_read(&netfs_n_rh_retry_read_subreq), + atomic_read(&netfs_n_wh_retry_write_req), + atomic_read(&netfs_n_wh_retry_write_subreq)); seq_printf(m, "Objs : rr=%u sr=%u foq=%u wsc=%u\n", atomic_read(&netfs_n_rh_rreq), atomic_read(&netfs_n_rh_sreq), diff --git a/fs/netfs/write_issue.c b/fs/netfs/write_issue.c index 69727411683e..77279fc5b5a7 100644 --- a/fs/netfs/write_issue.c +++ b/fs/netfs/write_issue.c @@ -253,6 +253,7 @@ void netfs_reissue_write(struct netfs_io_stream *stream, subreq->retry_count++; __clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags); __set_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags); + netfs_stat(&netfs_n_wh_retry_write_subreq); netfs_do_issue_write(stream, subreq); } diff --git a/fs/netfs/write_retry.c b/fs/netfs/write_retry.c index c841a851dd73..545d33079a77 100644 --- a/fs/netfs/write_retry.c +++ b/fs/netfs/write_retry.c @@ -203,6 +203,8 @@ void netfs_retry_writes(struct netfs_io_request *wreq) struct netfs_io_stream *stream; int s; + netfs_stat(&netfs_n_wh_retry_write_req); + /* Wait for all outstanding I/O to quiesce before performing retries as * we may need to renegotiate the I/O sizes. */ From 5de0219a9bb9dacc4ce6e8f2745540dcce786983 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 12 Feb 2025 22:24:01 +0000 Subject: [PATCH 5/5] netfs: Fix setting NETFS_RREQ_ALL_QUEUED to be after all subreqs queued Due to the code that queues a subreq on the active subrequest list getting moved to netfs_issue_read(), the NETFS_RREQ_ALL_QUEUED flag may now get set before the list-add actually happens. This is not a problem if the collection worker happens after the list-add, but it's a race - and, for 9P, where the read from the server is synchronous and done in the submitting thread, this is a lot more likely. The result is that, if the timing is wrong, a ref gets leaked because the collector thinks that all the subreqs have completed (because it can't see the last one yet) and clears NETFS_RREQ_IN_PROGRESS - at which point, the collection worker no longer goes into the collector. This can be provoked with AFS by injecting an msleep() right before the final subreq is queued. Fix this by splitting the queuing part out of netfs_issue_read() into a new function, netfs_queue_read(), and calling it separately. The setting of NETFS_RREQ_ALL_QUEUED is then done by netfs_queue_read() whilst it is holding the spinlock (that's probably unnecessary, but shouldn't hurt). It might be better to set a flag on the final subreq, but this could be a problem if an error occurs and we can't queue it. Fixes: e2d46f2ec332 ("netfs: Change the read result collector to only use one work item") Reported-by: Ihor Solodrai Closes: https://lore.kernel.org/r/a7x33d4dnMdGTtRivptq6S1i8btK70SNBP2XyX_xwDAhLvgQoPox6FVBOkifq4eBinfFfbZlIkMZBe3QarlWTxoEtHZwJCZbNKtaqrR7PvI=@pm.me/ Signed-off-by: David Howells Link: https://lore.kernel.org/r/20250212222402.3618494-4-dhowells@redhat.com Tested-by: Ihor Solodrai cc: Eric Van Hensbergen cc: Latchesar Ionkov cc: Dominique Martinet cc: Christian Schoenebeck cc: Marc Dionne cc: Steve French cc: Paulo Alcantara cc: Jeff Layton cc: v9fs@lists.linux.dev cc: linux-cifs@vger.kernel.org cc: netfs@lists.linux.dev cc: linux-fsdevel@vger.kernel.org Signed-off-by: Christian Brauner --- fs/netfs/buffered_read.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) diff --git a/fs/netfs/buffered_read.c b/fs/netfs/buffered_read.c index f761d44b3436..0d1b6d35ff3b 100644 --- a/fs/netfs/buffered_read.c +++ b/fs/netfs/buffered_read.c @@ -155,8 +155,9 @@ static void netfs_read_cache_to_pagecache(struct netfs_io_request *rreq, netfs_cache_read_terminated, subreq); } -static void netfs_issue_read(struct netfs_io_request *rreq, - struct netfs_io_subrequest *subreq) +static void netfs_queue_read(struct netfs_io_request *rreq, + struct netfs_io_subrequest *subreq, + bool last_subreq) { struct netfs_io_stream *stream = &rreq->io_streams[0]; @@ -177,8 +178,17 @@ static void netfs_issue_read(struct netfs_io_request *rreq, } } - spin_unlock(&rreq->lock); + if (last_subreq) { + smp_wmb(); /* Write lists before ALL_QUEUED. */ + set_bit(NETFS_RREQ_ALL_QUEUED, &rreq->flags); + } + spin_unlock(&rreq->lock); +} + +static void netfs_issue_read(struct netfs_io_request *rreq, + struct netfs_io_subrequest *subreq) +{ switch (subreq->source) { case NETFS_DOWNLOAD_FROM_SERVER: rreq->netfs_ops->issue_read(subreq); @@ -293,11 +303,8 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq) } size -= slice; start += slice; - if (size <= 0) { - smp_wmb(); /* Write lists before ALL_QUEUED. */ - set_bit(NETFS_RREQ_ALL_QUEUED, &rreq->flags); - } + netfs_queue_read(rreq, subreq, size <= 0); netfs_issue_read(rreq, subreq); cond_resched(); } while (size > 0);