io_uring-5.10-2020-10-20

-----BEGIN PGP SIGNATURE-----
 
 iQJEBAABCAAuFiEEwPw5LcreJtl1+l5K99NY+ylx4KYFAl+O9WEQHGF4Ym9lQGtl
 cm5lbC5kawAKCRD301j7KXHgpqajEADSD5PiO94YWTtVNWFUjn5RW+GlCE70/0VV
 XImdasmvM8nb48E+z2EW0Ky4vKXeVy5r+WZAeIYqPUHy/ogQDVpEn00NL7tFQmOz
 8UYrlZ3LLE/bSeWM5iavgG7TldVs/ZfspJ0hj3/Ac7jJpzuRGEI5TClsxJ0mWV39
 b2qT4OYBDhwvwVPZ/qhWgEwJXJFzFywckouIbMw8gPkveebUYeyu/yuScNGwYuiQ
 46YPEk/XIuOy8iUvQjqhLY+NNlAKJwt3z9WZgt5F+TZIhkpp7z6h20+jezFQcuFP
 GXzIDN+EADpsbw7MWJYIZVffxDEMlDpkJlAVMT1hsYLDfTXoEzmFwRddoFh2Fjf6
 ghWqhOKffUuAOX2xs1MrS2xLaxd0ot7QqZJVTYk7zEljkaRANlstSZZ+PpI+Sad/
 rNieQvs6jnsmTODDEaV3qyFX5aBQ2NdvyndZNU9wz0GZAWAdz+wxE0A1FVD0A37i
 p6m8sIvhNg3/cW89G04IDYUkAygi8knVDnEDHRwaJtswZQ4pRSGMp+N4qZ0GpnK7
 BviaAhofGaYlqruavO6Ug2YyomYpWGlUxTaB9ZKh0HkEDlDM945+0sgQRdxfsE8d
 OboycqJn3puOl/wh5Fc4oGYrWLsDbaA/5kksC4lm85Z+HUf+UXMS4QFdoPJYjhuM
 H6oMz1w2bQ==
 =v56S
 -----END PGP SIGNATURE-----

Merge tag 'io_uring-5.10-2020-10-20' of git://git.kernel.dk/linux-block

Pull io_uring updates from Jens Axboe:
 "A mix of fixes and a few stragglers. In detail:

   - Revert the bogus __read_mostly that we discussed for the initial
     pull request.

   - Fix a merge window regression with fixed file registration error
     path handling.

   - Fix io-wq numa node affinities.

   - Series abstracting out an io_identity struct, making it both easier
     to see what the personality items are, and also easier to to adopt
     more. Use this to cover audit logging.

   - Fix for read-ahead disabled block condition in async buffered
     reads, and using single page read-ahead to unify what
     generic_file_buffer_read() path is used.

   - Series for REQ_F_COMP_LOCKED fix and removal of it (Pavel)

   - Poll fix (Pavel)"

* tag 'io_uring-5.10-2020-10-20' of git://git.kernel.dk/linux-block: (21 commits)
  io_uring: use blk_queue_nowait() to check if NOWAIT supported
  mm: use limited read-ahead to satisfy read
  mm: mark async iocb read as NOWAIT once some data has been copied
  io_uring: fix double poll mask init
  io-wq: inherit audit loginuid and sessionid
  io_uring: use percpu counters to track inflight requests
  io_uring: assign new io_identity for task if members have changed
  io_uring: store io_identity in io_uring_task
  io_uring: COW io_identity on mismatch
  io_uring: move io identity items into separate struct
  io_uring: rely solely on work flags to determine personality.
  io_uring: pass required context in as flags
  io-wq: assign NUMA node locality if appropriate
  io_uring: fix error path cleanup in io_sqe_files_register()
  Revert "io_uring: mark io_uring_fops/io_op_defs as __read_mostly"
  io_uring: fix REQ_F_COMP_LOCKED by killing it
  io_uring: dig out COMP_LOCK from deep call chain
  io_uring: don't put a poll req under spinlock
  io_uring: don't unnecessarily clear F_LINK_TIMEOUT
  io_uring: don't set COMP_LOCKED if won't put
  ...

fs/io-wq.c

@@ -18,6 +18,7 @@
 #include <linux/fs_struct.h>
 #include <linux/task_work.h>
 #include <linux/blk-cgroup.h>
+#include <linux/audit.h>
 
 #include "io-wq.h"
 
@@ -429,14 +430,10 @@ static void io_wq_switch_mm(struct io_worker *worker, struct io_wq_work *work)
 		mmput(worker->mm);
 		worker->mm = NULL;
 	}
-	if (!work->mm)
-		return;
 
-	if (mmget_not_zero(work->mm)) {
-		kthread_use_mm(work->mm);
-		worker->mm = work->mm;
-		/* hang on to this mm */
-		work->mm = NULL;
+	if (mmget_not_zero(work->identity->mm)) {
+		kthread_use_mm(work->identity->mm);
+		worker->mm = work->identity->mm;
 		return;
 	}
 
@@ -448,9 +445,11 @@ static inline void io_wq_switch_blkcg(struct io_worker *worker,
 				      struct io_wq_work *work)
 {
 #ifdef CONFIG_BLK_CGROUP
-	if (work->blkcg_css != worker->blkcg_css) {
-		kthread_associate_blkcg(work->blkcg_css);
-		worker->blkcg_css = work->blkcg_css;
+	if (!(work->flags & IO_WQ_WORK_BLKCG))
+		return;
+	if (work->identity->blkcg_css != worker->blkcg_css) {
+		kthread_associate_blkcg(work->identity->blkcg_css);
+		worker->blkcg_css = work->identity->blkcg_css;
 	}
 #endif
 }
@@ -458,9 +457,9 @@ static inline void io_wq_switch_blkcg(struct io_worker *worker,
 static void io_wq_switch_creds(struct io_worker *worker,
 			       struct io_wq_work *work)
 {
-	const struct cred *old_creds = override_creds(work->creds);
+	const struct cred *old_creds = override_creds(work->identity->creds);
 
-	worker->cur_creds = work->creds;
+	worker->cur_creds = work->identity->creds;
 	if (worker->saved_creds)
 		put_cred(old_creds); /* creds set by previous switch */
 	else
@@ -470,20 +469,26 @@ static void io_wq_switch_creds(struct io_worker *worker,
 static void io_impersonate_work(struct io_worker *worker,
 				struct io_wq_work *work)
 {
-	if (work->files && current->files != work->files) {
+	if ((work->flags & IO_WQ_WORK_FILES) &&
+	    current->files != work->identity->files) {
 		task_lock(current);
-		current->files = work->files;
-		current->nsproxy = work->nsproxy;
+		current->files = work->identity->files;
+		current->nsproxy = work->identity->nsproxy;
 		task_unlock(current);
 	}
-	if (work->fs && current->fs != work->fs)
-		current->fs = work->fs;
-	if (work->mm != worker->mm)
+	if ((work->flags & IO_WQ_WORK_FS) && current->fs != work->identity->fs)
+		current->fs = work->identity->fs;
+	if ((work->flags & IO_WQ_WORK_MM) && work->identity->mm != worker->mm)
 		io_wq_switch_mm(worker, work);
-	if (worker->cur_creds != work->creds)
+	if ((work->flags & IO_WQ_WORK_CREDS) &&
+	    worker->cur_creds != work->identity->creds)
 		io_wq_switch_creds(worker, work);
-	current->signal->rlim[RLIMIT_FSIZE].rlim_cur = work->fsize;
+	current->signal->rlim[RLIMIT_FSIZE].rlim_cur = work->identity->fsize;
 	io_wq_switch_blkcg(worker, work);
+#ifdef CONFIG_AUDIT
+	current->loginuid = work->identity->loginuid;
+	current->sessionid = work->identity->sessionid;
+#endif
 }
 
 static void io_assign_current_work(struct io_worker *worker,
@@ -496,6 +501,11 @@ static void io_assign_current_work(struct io_worker *worker,
 		cond_resched();
 	}
 
+#ifdef CONFIG_AUDIT
+	current->loginuid = KUIDT_INIT(AUDIT_UID_UNSET);
+	current->sessionid = AUDIT_SID_UNSET;
+#endif
+
 	spin_lock_irq(&worker->lock);
 	worker->cur_work = work;
 	spin_unlock_irq(&worker->lock);
@@ -676,6 +686,7 @@ static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
 		kfree(worker);
 		return false;
 	}
+	kthread_bind_mask(worker->task, cpumask_of_node(wqe->node));
 
 	raw_spin_lock_irq(&wqe->lock);
 	hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);

fs/io-wq.h

@@ -1,6 +1,8 @@
 #ifndef INTERNAL_IO_WQ_H
 #define INTERNAL_IO_WQ_H
 
+#include <linux/io_uring.h>
+
 struct io_wq;
 
 enum {
@@ -10,6 +12,12 @@ enum {
 	IO_WQ_WORK_NO_CANCEL	= 8,
 	IO_WQ_WORK_CONCURRENT	= 16,
 
+	IO_WQ_WORK_FILES	= 32,
+	IO_WQ_WORK_FS		= 64,
+	IO_WQ_WORK_MM		= 128,
+	IO_WQ_WORK_CREDS	= 256,
+	IO_WQ_WORK_BLKCG	= 512,
+
 	IO_WQ_HASH_SHIFT	= 24,	/* upper 8 bits are used for hash key */
 };
 
@@ -85,15 +93,7 @@ static inline void wq_list_del(struct io_wq_work_list *list,
 
 struct io_wq_work {
 	struct io_wq_work_node list;
-	struct files_struct *files;
-	struct mm_struct *mm;
-#ifdef CONFIG_BLK_CGROUP
-	struct cgroup_subsys_state *blkcg_css;
-#endif
-	const struct cred *creds;
-	struct nsproxy *nsproxy;
-	struct fs_struct *fs;
-	unsigned long fsize;
+	struct io_identity *identity;
 	unsigned flags;
 };
 

fs/proc/base.c

@@ -1268,6 +1268,10 @@ static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
 	kuid_t kloginuid;
 	int rv;
 
+	/* Don't let kthreads write their own loginuid */
+	if (current->flags & PF_KTHREAD)
+		return -EPERM;
+
 	rcu_read_lock();
 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
 		rcu_read_unlock();

include/linux/io_uring.h

@@ -4,18 +4,33 @@
 
 #include <linux/sched.h>
 #include <linux/xarray.h>
-#include <linux/percpu-refcount.h>
+
+struct io_identity {
+	struct files_struct		*files;
+	struct mm_struct		*mm;
+#ifdef CONFIG_BLK_CGROUP
+	struct cgroup_subsys_state	*blkcg_css;
+#endif
+	const struct cred		*creds;
+	struct nsproxy			*nsproxy;
+	struct fs_struct		*fs;
+	unsigned long			fsize;
+#ifdef CONFIG_AUDIT
+	kuid_t				loginuid;
+	unsigned int			sessionid;
+#endif
+	refcount_t			count;
+};
 
 struct io_uring_task {
 	/* submission side */
 	struct xarray		xa;
 	struct wait_queue_head	wait;
 	struct file		*last;
-	atomic_long_t		req_issue;
-
-	/* completion side */
-	bool			in_idle ____cacheline_aligned_in_smp;
-	atomic_long_t		req_complete;
+	struct percpu_counter	inflight;
+	struct io_identity	__identity;
+	struct io_identity	*identity;
+	bool			in_idle;
 };
 
 #if defined(CONFIG_IO_URING)

mm/filemap.c

@@ -2199,6 +2199,14 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 	last_index = (*ppos + iter->count + PAGE_SIZE-1) >> PAGE_SHIFT;
 	offset = *ppos & ~PAGE_MASK;
 
+	/*
+	 * If we've already successfully copied some data, then we
+	 * can no longer safely return -EIOCBQUEUED. Hence mark
+	 * an async read NOWAIT at that point.
+	 */
+	if (written && (iocb->ki_flags & IOCB_WAITQ))
+		iocb->ki_flags |= IOCB_NOWAIT;
+
 	for (;;) {
 		struct page *page;
 		pgoff_t end_index;

mm/readahead.c

@@ -552,15 +552,23 @@ readit:
 void page_cache_sync_ra(struct readahead_control *ractl,
 		struct file_ra_state *ra, unsigned long req_count)
 {
-	/* no read-ahead */
-	if (!ra->ra_pages)
-		return;
+	bool do_forced_ra = ractl->file && (ractl->file->f_mode & FMODE_RANDOM);
 
-	if (blk_cgroup_congested())
-		return;
+	/*
+	 * Even if read-ahead is disabled, issue this request as read-ahead
+	 * as we'll need it to satisfy the requested range. The forced
+	 * read-ahead will do the right thing and limit the read to just the
+	 * requested range, which we'll set to 1 page for this case.
+	 */
+	if (!ra->ra_pages || blk_cgroup_congested()) {
+		if (!ractl->file)
+			return;
+		req_count = 1;
+		do_forced_ra = true;
+	}
 
 	/* be dumb */
-	if (ractl->file && (ractl->file->f_mode & FMODE_RANDOM)) {
+	if (do_forced_ra) {
 		force_page_cache_ra(ractl, ra, req_count);
 		return;
 	}