Re: [PATCH v2] HID: cp2112: support large i2c transfers in hid-cp2112
From: Vaibhav Hiremath <hidden>
Date: 2015-07-08 12:07:07
Also in:
linux-i2c
On Wednesday 08 July 2015 03:55 PM, Ellen Wang wrote:
quoted hunk
cp2112_i2c_xfer() only reads up to 61 bytes, returning EIO on longers reads. The fix is to wrap a loop around cp2112_read() to pick up all the returned data. Signed-off-by: Ellen Wang <redacted> --- This is the updated patch with a check for 0 return from cp2112_read(). I tested it with a suitable delay in the loop to trigger the cp2112_raw_event() overrun bug, which must be fixed before this patch is applied. --- drivers/hid/hid-cp2112.c | 33 ++++++++++++++++++++++++++------- 1 file changed, 26 insertions(+), 7 deletions(-)diff --git a/drivers/hid/hid-cp2112.c b/drivers/hid/hid-cp2112.c index 3318de6..e2ffac0 100644 --- a/drivers/hid/hid-cp2112.c +++ b/drivers/hid/hid-cp2112.c@@ -509,13 +509,32 @@ static int cp2112_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, if (!(msgs->flags & I2C_M_RD)) goto finish; - ret = cp2112_read(dev, msgs->buf, msgs->len); - if (ret < 0) - goto power_normal; - if (ret != msgs->len) { - hid_warn(hdev, "short read: %d < %d\n", ret, msgs->len); - ret = -EIO; - goto power_normal; + for (count = 0; count < msgs->len;) { + ret = cp2112_read(dev, msgs->buf + count, msgs->len - count); + hid_warn(hdev, "read returned %d for %zd\n", + ret, msgs->len - count);
Do you always want to throw warning here, unconditionally ?
+ if (ret < 0)
+ goto power_normal;
+ if (ret == 0) {
+ hid_err(hdev, "read returned 0\n");
+ ret = -EIO;
+ goto power_normal;
+ }
bit simplified, I guess :)
if (ret < 0 || ret == 0) {
hid_err(hdev, "read returned %d", ret);
ret = ret == 0 ? -EIO : ret;
goto power_normal;
}
+ count += ret;
+ if (count > msgs->len) {
+ /*
+ * The hardware returned too much data.
+ * This is mostly harmless because cp2112_read()
+ * has a limit check so didn't overrun our
+ * buffer. Nevertheless, we return an error
+ * because something is seriously wrong and
+ * it shouldn't go unnoticed.
+ */
+ hid_err(hdev, "long read: %d > %zd\n",
+ ret, msgs->len - count + ret);You may want to take another look here. 'ret' will be either, - ret = msgs->len Not applicable - ret > msgs->len (count > msgs->len) will happen in one single iteration, and will - ret < msgs->len (count > msgs->len) will happen in multiple iterations where count keeps incrementing based on ret In the 2 scenarios above, I believe you would want to show, actual read bytes > requested read bytes Am I missing something here? Thanks, Vaibhav