linux-kernel - [PATCH 5.4 001/152] io_uring: grab ->fs as part of async offload

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue,  3 Mar 2020 18:41:39 +0100
From:   Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To:     linux-kernel@...r.kernel.org
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        stable@...r.kernel.org, Jens Axboe <axboe@...nel.dk>,
        Sasha Levin <sashal@...nel.org>
Subject: [PATCH 5.4 001/152] io_uring: grab ->fs as part of async offload

From: Jens Axboe <axboe@...nel.dk>

[ Upstream commits 9392a27d88b9 and ff002b30181d ]

Ensure that the async work grabs ->fs from the queueing task if the
punted commands needs to do lookups.

We don't have these two commits in 5.4-stable:

ff002b30181d30cdfbca316dadd099c3ca0d739c
9392a27d88b9707145d713654eb26f0c29789e50

because they don't apply with the rework that was done in how io_uring
handles offload. Since there's no io-wq in 5.4, it doesn't make sense to
do two patches. I'm attaching my port of the two for 5.4-stable, it's
been tested. Please queue it up for the next 5.4-stable, thanks!

Signed-off-by: Jens Axboe <axboe@...nel.dk>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
 fs/io_uring.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index ed9a551882cf3..f34a8f7eee5d1 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -71,6 +71,7 @@
 #include <linux/sizes.h>
 #include <linux/hugetlb.h>
 #include <linux/highmem.h>
+#include <linux/fs_struct.h>
 
 #include <uapi/linux/io_uring.h>
 
@@ -334,6 +335,8 @@ struct io_kiocb {
 	u32			result;
 	u32			sequence;
 
+	struct fs_struct	*fs;
+
 	struct work_struct	work;
 };
 
@@ -651,6 +654,7 @@ static struct io_kiocb *io_get_req(struct io_ring_ctx *ctx,
 	/* one is dropped after submission, the other at completion */
 	refcount_set(&req->refs, 2);
 	req->result = 0;
+	req->fs = NULL;
 	return req;
 out:
 	percpu_ref_put(&ctx->refs);
@@ -1663,6 +1667,16 @@ static int io_send_recvmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 			ret = -EINTR;
 	}
 
+	if (req->fs) {
+		struct fs_struct *fs = req->fs;
+
+		spin_lock(&req->fs->lock);
+		if (--fs->users)
+			fs = NULL;
+		spin_unlock(&req->fs->lock);
+		if (fs)
+			free_fs_struct(fs);
+	}
 	io_cqring_add_event(req->ctx, sqe->user_data, ret);
 	io_put_req(req);
 	return 0;
@@ -2159,6 +2173,7 @@ static inline bool io_sqe_needs_user(const struct io_uring_sqe *sqe)
 static void io_sq_wq_submit_work(struct work_struct *work)
 {
 	struct io_kiocb *req = container_of(work, struct io_kiocb, work);
+	struct fs_struct *old_fs_struct = current->fs;
 	struct io_ring_ctx *ctx = req->ctx;
 	struct mm_struct *cur_mm = NULL;
 	struct async_list *async_list;
@@ -2178,6 +2193,15 @@ restart:
 		/* Ensure we clear previously set non-block flag */
 		req->rw.ki_flags &= ~IOCB_NOWAIT;
 
+		if (req->fs != current->fs && current->fs != old_fs_struct) {
+			task_lock(current);
+			if (req->fs)
+				current->fs = req->fs;
+			else
+				current->fs = old_fs_struct;
+			task_unlock(current);
+		}
+
 		ret = 0;
 		if (io_sqe_needs_user(sqe) && !cur_mm) {
 			if (!mmget_not_zero(ctx->sqo_mm)) {
@@ -2276,6 +2300,11 @@ out:
 		mmput(cur_mm);
 	}
 	revert_creds(old_cred);
+	if (old_fs_struct) {
+		task_lock(current);
+		current->fs = old_fs_struct;
+		task_unlock(current);
+	}
 }
 
 /*
@@ -2503,6 +2532,23 @@ err:
 
 	req->user_data = s->sqe->user_data;
 
+#if defined(CONFIG_NET)
+	switch (READ_ONCE(s->sqe->opcode)) {
+	case IORING_OP_SENDMSG:
+	case IORING_OP_RECVMSG:
+		spin_lock(&current->fs->lock);
+		if (!current->fs->in_exec) {
+			req->fs = current->fs;
+			req->fs->users++;
+		}
+		spin_unlock(&current->fs->lock);
+		if (!req->fs) {
+			ret = -EAGAIN;
+			goto err_req;
+		}
+	}
+#endif
+
 	/*
 	 * If we already have a head request, queue this one for async
 	 * submittal once the head completes. If we don't have a head but
-- 
2.20.1