[PATCH] usb: ffs: fix regression when quirk_ep_out_aligned

Discussion:

[PATCH] usb: ffs: fix regression when quirk_ep_out_aligned_size flag is set

David Cohen

2014-10-08 00:18:06 UTC

Permalink

--
2.1.0

Felipe Balbi

2014-10-08 00:31:44 UTC

Permalink

Hi,

Post by David Cohen
The commit '2e4c7553cd usb: gadget: f_fs: add aio support' broke the

did that commit non-aio or is only aio broken ?

Post by David Cohen
quirk implemented to align buffer size to maxpacketsize on out endpoint.
As result, functionfs does not work on Intel platforms using dwc3 driver
(i.e. Bay Trail and Merrifield). This patch fixes the issue.

and Braswell.

Post by David Cohen
This code is based on a previous Qiuxu's patch.

How have you tested this ? Do you have a test-case to share ? Then I can
add it to my tests which I have been running on my platforms too (they
include USB20CV and USB30CV where applicable - quite a few fixes coming
soon).

cheers

--
balbi

David Cohen

2014-10-08 17:53:52 UTC

Permalink

Post by Felipe Balbi
Hi,

Hi Felipe,

Post by Felipe Balbi

Post by David Cohen
The commit '2e4c7553cd usb: gadget: f_fs: add aio support' broke the

did that commit non-aio or is only aio broken ?

That commit broke the quirk on both cases.

Post by Felipe Balbi

and Braswell.

Post by David Cohen
This code is based on a previous Qiuxu's patch.

My testcase is to use Android's adbd service via function_fs (no
out-of-tree android gadget). Without the quitk, it doesn't work.

Br, David

Post by Felipe Balbi
cheers
--
balbi

Felipe Balbi

2014-10-08 00:32:56 UTC

Permalink

Hi again,

btw, please resend and add below right here:

Fixes: 2e4c7553cd (usb: gadget: f_fs: add aio support)

--
balbi

David Cohen

2014-10-08 17:55:58 UTC

Permalink

Post by Felipe Balbi
Hi again,

Fixes: 2e4c7553cd (usb: gadget: f_fs: add aio support)

Sure. But v3.16 doesn't have the new 'function' directory under usb/gadget.
Not sure if same patch is useful for v3.16+.
The patch I sent is intended for v3.17.1 and v3.18-rc1.

Br, David

Post by Felipe Balbi
--
balbi

Felipe Balbi

2014-10-08 19:54:54 UTC

Permalink

Hi,

Post by David Cohen

Post by Felipe Balbi
Hi again,

Fixes: 2e4c7553cd (usb: gadget: f_fs: add aio support)

Sure. But v3.16 doesn't have the new 'function' directory under usb/gadget.
Not sure if same patch is useful for v3.16+.
The patch I sent is intended for v3.17.1 and v3.18-rc1.

code is the same anyway right ? It's just a path change :-)

--
balbi

Michal Nazarewicz

2014-10-08 11:34:16 UTC

Permalink

Post by David Cohen
The commit '2e4c7553cd usb: gadget: f_fs: add aio support' broke the
quirk implemented to align buffer size to maxpacketsize on out endpoi=

nt.

Post by David Cohen
As result, functionfs does not work on Intel platforms using dwc3 dri=

ver

Post by David Cohen
(i.e. Bay Trail and Merrifield). This patch fixes the issue.
This code is based on a previous Qiuxu's patch.
---
Hi,
Since this is a feature that worked in past, this patch is meant for =

3.17

Post by David Cohen
kernel. If/when we pass review and accept it on 3.17, I'll send a ver=

sion for

Post by David Cohen
stable 3.16 kernel too. It is not required for 3.14, since the regres=

sion came

Post by David Cohen
later.
Br, David Cohen
---
drivers/usb/gadget/function/f_fs.c | 21 ++++++++++++++++-----
1 file changed, 16 insertions(+), 5 deletions(-)
diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/=

function/f_fs.c

Post by David Cohen
index 0dc3552d1360..6e2b8063b170 100644
--- a/drivers/usb/gadget/function/f_fs.c
+++ b/drivers/usb/gadget/function/f_fs.c
@@ -648,15 +648,26 @@ static void ffs_user_copy_worker(struct work_st=

ruct *work)

Post by David Cohen
if (io_data->read && ret > 0) {
int i;
size_t pos =3D 0;
+
+ /*
+ * Since req->length may be bigger than io_data->len (after
+ * being rounded up to maxpacketsize), we may end up with more
+ * data then user space has space for.
+ */
+ ret =3D min_t(int, ret, io_data->len);
+
use_mm(io_data->mm);
for (i =3D 0; i < io_data->nr_segs; i++) {
+ size_t len =3D min_t(size_t, ret - pos,
+ io_data->iovec[i].iov_len);
+ if (!len)
+ break;
if (unlikely(copy_to_user(io_data->iovec[i].iov_base,
- &io_data->buf[pos],
- io_data->iovec[i].iov_len))) {
+ &io_data->buf[pos], len))) {
ret =3D -EFAULT;
break;
}
- pos +=3D io_data->iovec[i].iov_len;
+ pos +=3D len;
}
unuse_mm(io_data->mm);
}
@@ -794,7 +805,7 @@ static ssize_t ffs_epfile_io(struct file *file, s=

truct ffs_io_data *io_data)

Post by David Cohen
goto error_lock;
=20
req->buf =3D data;
- req->length =3D io_data->len;
+ req->length =3D data_len;
=20
io_data->buf =3D data;
io_data->ep =3D ep->ep;
@@ -816,7 +827,7 @@ static ssize_t ffs_epfile_io(struct file *file, s=

truct ffs_io_data *io_data)

Post by David Cohen
=20
req =3D ep->req;
req->buf =3D data;
- req->length =3D io_data->len;
+ req->length =3D data_len;
=20
req->context =3D &done;
req->complete =3D ffs_epfile_io_complete;
--=20
2.1.0

--=20
Best regards, _ _
=2Eo. | Liege of Serenely Enlightened Majesty of o' \,=3D./ `o
=2E.o | Computer Science, Micha=C5=82 =E2=80=9Cmina86=E2=80=9D Nazarew=
icz (o o)
ooo +--<***@google.com>--<xmpp:***@jabber.org>--ooO--(_)--Ooo--

David Cohen

2014-10-08 21:12:18 UTC

Permalink

--
2.1.0

Al Viro

2014-10-12 19:12:28 UTC

Permalink

Post by David Cohen
use_mm(io_data->mm);
for (i = 0; i < io_data->nr_segs; i++) {
+ size_t len = min_t(size_t, ret - pos,
+ io_data->iovec[i].iov_len);
+ if (!len)
+ break;
if (unlikely(copy_to_user(io_data->iovec[i].iov_base,
- &io_data->buf[pos],
- io_data->iovec[i].iov_len))) {
+ &io_data->buf[pos], len))) {
ret = -EFAULT;
break;
}
- pos += io_data->iovec[i].iov_len;
+ pos += len;

Hmm... This is really asking for something like
if (copy_to_iter(io_data->buf, ret, <something>) != ret)
ret = -EFAULT;
with <something> being an iov_iter instead of iovec. It would be really
nice to have that thing switched to ->read_iter/->write_iter, dropping
->read/->write/->aio_read/->aio_write; I'm not familiar enough with that
code to do it on my own, though, so it would require some help from
maintainers...

Felipe Balbi

2014-10-12 19:58:57 UTC

Permalink

Post by Al Viro

cool, Michal, if you're too busy lately, I can look at that one myself.
I suppose the test application we have under tool/usb is good enough for
validation ?

cheers

--
balbi

Felipe Balbi

2014-10-13 15:32:12 UTC

Permalink

Post by David Cohen
The commit '2e4c7553cd usb: gadget: f_fs: add aio support' broke the
quirk implemented to align buffer size to maxpacketsize on out endpoint.
As result, functionfs does not work on Intel platforms using dwc3 driver
(i.e. Bay Trail and Merrifield). This patch fixes the issue.
This code is based on a previous Qiuxu's patch.
Fixes: 2e4c7553cd (usb: gadget: f_fs: add aio support)
---
Hi,
Since this is a feature that worked in past, this is meant for stable
versions >= 3.16 too.
v1 to v2: just added Fixes, Cc and Acked-by lines on patch description.

this adds a build warning for use of maybe unitialized data_len. Plese
fix.

--
balbi

David Cohen

2014-10-13 16:55:56 UTC

Permalink

Post by Felipe Balbi

this adds a build warning for use of maybe unitialized data_len. Plese
fix.

It's a false-positive warning. data_len is only initialized if halt != 0
and it's only used if halt != 0 too.

Do you prefer to initialize it to 0 during the declaration to silent the
compiler?

BR, David

Post by Felipe Balbi
--
balbi

Felipe Balbi

2014-10-13 17:03:24 UTC

Permalink

Post by David Cohen

Post by Felipe Balbi

this adds a build warning for use of maybe unitialized data_len. Plese
fix.

It's a false-positive warning. data_len is only initialized if halt != 0
and it's only used if halt != 0 too.
Do you prefer to initialize it to 0 during the declaration to silent the
compiler?

yup.

--
balbi

David Cohen

2014-10-13 18:15:54 UTC

Permalink

The commit '2e4c7553cd usb: gadget: f_fs: add aio support' broke the
quirk implemented to align buffer size to maxpacketsize on out endpoint.
As result, functionfs does not work on Intel platforms using dwc3 driver
(i.e. Bay Trail and Merrifield). This patch fixes the issue.

This code is based on a previous Qiuxu's patch.

Fixes: 2e4c7553cd (usb: gadget: f_fs: add aio support)
Cc: <***@vger.kernel.org> # v3.16+
Signed-off-by: David Cohen <***@linux.intel.com>
Signed-off-by: Qiuxu Zhuo <***@intel.com>
Acked-by: Michal Nazarewicz <***@mina86.com>
---

Hi,

Since this is a feature that worked in past, this is meant for stable
versions >= 3.16 too.

v1 to v2: just added Fixes, Cc and Acked-by lines on patch description.
v2 to v3: fixed compiler warning about data_len being used unitialized

Br, David Cohen

---
drivers/usb/gadget/function/f_fs.c | 40 ++++++++++++++++++++++++++++++++------
1 file changed, 34 insertions(+), 6 deletions(-)

diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/function/f_fs.c
index 0dc3552d1360..9b6bc4d30352 100644
--- a/drivers/usb/gadget/function/f_fs.c
+++ b/drivers/usb/gadget/function/f_fs.c
@@ -648,15 +648,26 @@ static void ffs_user_copy_worker(struct work_struct *work)
if (io_data->read && ret > 0) {
int i;
size_t pos = 0;
+
+ /*
+ * Since req->length may be bigger than io_data->len (after
+ * being rounded up to maxpacketsize), we may end up with more
+ * data then user space has space for.
+ */
+ ret = min_t(int, ret, io_data->len);
+
use_mm(io_data->mm);
for (i = 0; i < io_data->nr_segs; i++) {
+ size_t len = min_t(size_t, ret - pos,
+ io_data->iovec[i].iov_len);
+ if (!len)
+ break;
if (unlikely(copy_to_user(io_data->iovec[i].iov_base,
- &io_data->buf[pos],
- io_data->iovec[i].iov_len))) {
+ &io_data->buf[pos], len))) {
ret = -EFAULT;
break;
}
- pos += io_data->iovec[i].iov_len;
+ pos += len;
}
unuse_mm(io_data->mm);
}
@@ -688,7 +699,7 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
struct ffs_epfile *epfile = file->private_data;
struct ffs_ep *ep;
char *data = NULL;
- ssize_t ret, data_len;
+ ssize_t ret, data_len = -EINVAL;
int halt;

/* Are we still active? */
@@ -788,13 +799,30 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
/* Fire the request */
struct usb_request *req;

+ /*
+ * Sanity Check: even though data_len can't be used
+ * uninitialized at the time I write this comment, some
+ * compilers complain about this situation.
+ * In order to keep the code clean from warnings, data_len is
+ * being initialized to -EINVAL during its declaration, which
+ * means we can't rely on compiler anymore to warn no future
+ * changes won't result in data_len being used uninitialized.
+ * For such reason, we're adding this redundant sanity check
+ * here.
+ */
+ if (unlikely(data_len == -EINVAL)) {
+ WARN(1, "%s: data_len == -EINVAL\n", __func__);
+ ret = -EINVAL;
+ goto error_lock;
+ }
+
if (io_data->aio) {
req = usb_ep_alloc_request(ep->ep, GFP_KERNEL);
if (unlikely(!req))
goto error_lock;

req->buf = data;
- req->length = io_data->len;
+ req->length = data_len;

io_data->buf = data;
io_data->ep = ep->ep;
@@ -816,7 +844,7 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)

req = ep->req;
req->buf = data;
- req->length = io_data->len;
+ req->length = data_len;

req->context = &done;
req->complete = ffs_epfile_io_complete;

--
2.1.0