[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <e3987a2f55d154a7b217d86d805719043957db60.camel@fb.com>
Date: Thu, 21 Jul 2022 12:05:49 +0000
From: Dylan Yudaken <dylany@...com>
To: "ammarfaizi2@...weeb.org" <ammarfaizi2@...weeb.org>,
"axboe@...nel.dk" <axboe@...nel.dk>
CC: "gwml@...r.gnuweeb.org" <gwml@...r.gnuweeb.org>,
"fernandafmr12@...weeb.org" <fernandafmr12@...weeb.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"asml.silence@...il.com" <asml.silence@...il.com>,
"io-uring@...r.kernel.org" <io-uring@...r.kernel.org>
Subject: Re: Linux 5.19-rc7 liburing test `poll-mshot-overflow.t` and
`read-write.t` fail
On Thu, 2022-07-21 at 17:41 +0700, Ammar Faizi wrote:
> On 7/21/22 4:48 PM, Dylan Yudaken wrote:
> > What fs are you using? testing on a fresh XFS fs read-write.t works
> > for
> > me
>
> I am using btrfs.
>
> After I got your email, I tried to run the test on an ext4 directory
> and
> it works fine. But fails on a btrfs directory.
>
> Any idea why does the test fail on a btrfs fs?
>
It seems to be a problem with blocking reads, buffer select and READV.
My guess is that ext4/xfs are not blocking.
in b66e65f41426 ("io_uring: never call io_buffer_select() for a buffer
re-select"), this line was added in __io_iov_buffer_select
- iov[0].iov_len = len;
+ req->rw.len = iov[0].iov_len = len;
Basically stashing the buffer length in rw.len. The problem is that the
next time around that breaks at
if (req->rw.len != 1)
return -EINVAL;
The below fixes it as an example, but it's not great. Maybe someone can
figure out a better patch? Otherwise I can try tomorrow:
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 2b7bb62c7805..d9fa226f8e30 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -647,6 +647,8 @@ struct io_rw {
u64 addr;
u32 len;
rwf_t flags;
+ u64 bufaddr;
+ u32 buflen;
};
struct io_connect {
@@ -3899,7 +3901,7 @@ static ssize_t io_compat_import(struct io_kiocb
*req, struct iovec *iov,
return -ENOBUFS;
req->rw.addr = (unsigned long) buf;
iov[0].iov_base = buf;
- req->rw.len = iov[0].iov_len = (compat_size_t) len;
+ req->rw.buflen = iov[0].iov_len = (compat_size_t) len;
return 0;
}
#endif
@@ -3920,9 +3922,9 @@ static ssize_t __io_iov_buffer_select(struct
io_kiocb *req, struct iovec *iov,
buf = io_buffer_select(req, &len, issue_flags);
if (!buf)
return -ENOBUFS;
- req->rw.addr = (unsigned long) buf;
+ req->rw.bufaddr = (unsigned long) buf;
iov[0].iov_base = buf;
- req->rw.len = iov[0].iov_len = len;
+ req->rw.buflen = iov[0].iov_len = len;
return 0;
}
@@ -3930,8 +3932,8 @@ static ssize_t io_iov_buffer_select(struct
io_kiocb *req, struct iovec *iov,
unsigned int issue_flags)
{
if (req->flags & (REQ_F_BUFFER_SELECTED|REQ_F_BUFFER_RING)) {
- iov[0].iov_base = u64_to_user_ptr(req->rw.addr);
- iov[0].iov_len = req->rw.len;
+ iov[0].iov_base = u64_to_user_ptr(req->rw.bufaddr);
+ iov[0].iov_len = req->rw.buflen;
return 0;
}
if (req->rw.len != 1)
Powered by blists - more mailing lists