Re: [syzbot] [io-uring?] BUG: unable to handle kernel NULL pointer dereference in __io_remove_buffers (2)

[Jens Axboe <axboe@kernel.dk>]

On 10/2/23 8:38 AM, syzbot wrote:
> Hello,
> 
> syzbot found the following issue on:
> 
> HEAD commit:    ec8c298121e3 Merge tag 'x86-urgent-2023-10-01' of git://gi..
> git tree:       upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=16ef0ed6680000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=3be743fa9361d5b0
> dashboard link: https://syzkaller.appspot.com/bug?extid=2113e61b8848fa7951d8
> compiler:       arm-linux-gnueabi-gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40
> userspace arch: arm

I tried the syz repro in the console output, but can't trigger it. It
also makes very little sense to me... For when there is a reproducer,
the below would perhaps shed some light on it. We have bl->is_mapped ==
1, yet bl->buf_ring is NULL. Probably some artifact of 32-bit arm?


diff --git a/io_uring/kbuf.c b/io_uring/kbuf.c
index 556f4df25b0f..d5133ac8005f 100644
--- a/io_uring/kbuf.c
+++ b/io_uring/kbuf.c
@@ -504,6 +504,9 @@ static int io_pin_pbuf_ring(struct io_uring_buf_reg *reg,
 		return -EINVAL;
 	}
 #endif
+	WARN_ON_ONCE(!pages);
+	WARN_ON_ONCE(!nr_pages);
+	WARN_ON_ONCE(!br);
 	bl->buf_pages = pages;
 	bl->buf_nr_pages = nr_pages;
 	bl->buf_ring = br;
diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
index d9c853d10587..7034be555334 100644
--- a/io_uring/rsrc.c
+++ b/io_uring/rsrc.c
@@ -1037,39 +1037,36 @@ struct page **io_pin_pages(unsigned long ubuf, unsigned long len, int *npages)
 {
 	unsigned long start, end, nr_pages;
 	struct page **pages = NULL;
-	int pret, ret = -ENOMEM;
+	int ret;
 
 	end = (ubuf + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
 	start = ubuf >> PAGE_SHIFT;
 	nr_pages = end - start;
+	WARN_ON(!nr_pages);
 
 	pages = kvmalloc_array(nr_pages, sizeof(struct page *), GFP_KERNEL);
 	if (!pages)
-		goto done;
+		return ERR_PTR(-ENOMEM);
 
-	ret = 0;
 	mmap_read_lock(current->mm);
-	pret = pin_user_pages(ubuf, nr_pages, FOLL_WRITE | FOLL_LONGTERM,
-			      pages);
-	if (pret == nr_pages)
+	ret = pin_user_pages(ubuf, nr_pages, FOLL_WRITE | FOLL_LONGTERM, pages);
+	mmap_read_unlock(current->mm);
+
+	/* success, mapped all pages */
+	if (ret == nr_pages) {
 		*npages = nr_pages;
-	else
-		ret = pret < 0 ? pret : -EFAULT;
+		return pages;
+	}
 
-	mmap_read_unlock(current->mm);
-	if (ret) {
+	/* partial map, or didn't map anything */
+	if (ret >= 0) {
 		/* if we did partial map, release any pages we did get */
-		if (pret > 0)
-			unpin_user_pages(pages, pret);
-		goto done;
-	}
-	ret = 0;
-done:
-	if (ret < 0) {
-		kvfree(pages);
-		pages = ERR_PTR(ret);
+		if (ret)
+			unpin_user_pages(pages, ret);
+		ret = -EFAULT;
 	}
-	return pages;
+	kvfree(pages);
+	return ERR_PTR(ret);
 }
 
 static int io_sqe_buffer_register(struct io_ring_ctx *ctx, struct iovec *iov,

-- 
Jens Axboe