Skip to contentSkip to conversation

rtl2832_sdr: move from staging to media

PATCH New Dan Carpenter 2 participants 3 messages May 13, 2015 – May 14, 2015

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

Hello Antti Palosaari,

The patch 77bbb2b049c1: "rtl2832_sdr: move from staging to media" from Jul 15, 2014, leads to the following static checker warning:

	drivers/media/dvb-frontends/rtl2832_sdr.c:1265 rtl2832_sdr_s_ctrl()
	warn: test_bit() bitwise op in bit number

This is harmless but messy.

drivers/media/dvb-frontends/rtl2832_sdr.c
109

   110  struct rtl2832_sdr_dev {
   111  #define POWER_ON           (1 << 1)
   112  #define URB_BUF            (1 << 2)

We were supposed to use these to set ->flags on the next line.

113 unsigned long flags;
114
115 struct platform_device *pdev;
116
117 struct video_device vdev;
118 struct v4l2_device v4l2_dev;
119

[ snip ]

389 dev_dbg(&pdev->dev, "alloc buf=%d %p (dma %llu)\n",
390 dev->buf_num, dev->buf_list[dev->buf_num],
391 (long long)dev->dma_addr[dev->buf_num]);

   392                  dev->flags |= USB_STATE_URB_BUF;
                                      ^^^^^^^^^^^^^^^^^

But we use USB_STATE_URB_BUF (0x1) instead of URB_BUF.

393 }

[ snip ]

1263 c->bandwidth_hz = dev->bandwidth->val;
1264
1265 if (!test_bit(POWER_ON, &dev->flags))

                                      ^^^^^^^^

The original intent of the code was we test "if (dev->flags & POWER_ON)" but really what this is doing is "if (dev->flags & (1 << POWER_ON))" which is fine because we do it consistently, but it's not pretty and it causes static checkers to complain (and rightfully so).

  1266                          return 0;
  1267  
  1268                  if (fe->ops.tuner_ops.set_params)
  1269                          ret = fe->ops.tuner_ops.set_params(fe);
  1270                  else
  1271                          ret = 0;
  1272                  break;
  1273          default:

regards,
dan carpenter

#1 lore raw

Moikka!

Quoted 48 lines · Dan Carpenter

Hello Antti Palosaari,

The patch 77bbb2b049c1: "rtl2832_sdr: move from staging to media" from Jul 15, 2014, leads to the following static checker warning:

	drivers/media/dvb-frontends/rtl2832_sdr.c:1265 rtl2832_sdr_s_ctrl()
	warn: test_bit() bitwise op in bit number

This is harmless but messy.

drivers/media/dvb-frontends/rtl2832_sdr.c

    109
    110  struct rtl2832_sdr_dev {
    111  #define POWER_ON           (1 << 1)
    112  #define URB_BUF            (1 << 2)

We were supposed to use these to set ->flags on the next line.

    113          unsigned long flags;
    114
    115          struct platform_device *pdev;
    116
    117          struct video_device vdev;
    118          struct v4l2_device v4l2_dev;
    119

[ snip ]

    389                  dev_dbg(&pdev->dev, "alloc buf=%d %p (dma %llu)\n",
    390                          dev->buf_num, dev->buf_list[dev->buf_num],
    391                          (long long)dev->dma_addr[dev->buf_num]);
    392                  dev->flags |= USB_STATE_URB_BUF;
                                       ^^^^^^^^^^^^^^^^^

But we use USB_STATE_URB_BUF (0x1) instead of URB_BUF.

    393          }

[ snip ]

   1263                  c->bandwidth_hz = dev->bandwidth->val;
   1264
   1265                  if (!test_bit(POWER_ON, &dev->flags))
                                       ^^^^^^^^

The original intent of the code was we test "if (dev->flags & POWER_ON)" but really what this is doing is "if (dev->flags & (1 << POWER_ON))" which is fine because we do it consistently, but it's not pretty and it causes static checkers to complain (and rightfully so).

Quoted 9 lines
   1266                          return 0;
   1267
   1268                  if (fe->ops.tuner_ops.set_params)
   1269                          ret = fe->ops.tuner_ops.set_params(fe);
   1270                  else
   1271                          ret = 0;
   1272                  break;
   1273          default:

If you wish, you could fix those. Otherwise I will check issues pointed and correct. Lets say I will wait at least one week your patch.

[At the some point I am going to rewrote that USB streaming code as I am not happy with it. Also I have one device which needs to stream data both ways, from device to computer and from computer to device, which should be take into account.]

regards
Antti

Signature · 1 line
http://palosaari.fi/
#2 lore raw

PATCH 1/1 rtl2832_sdr: cleanup some set_bit() calls

This code works fine but static checkers complain. The test_bit() function takes the bit number and not a mask. Then the other issue is that we were using USB_STATE_URB_BUF which is BIT(0) instead of URB_BUF. Also we were open coding that instead of using the test/clear/set_bit() functions.

drivers/media/dvb-frontends/rtl2832_sdr.c +5 -5
index 3ff8806..6362c6c 100644
@@ -108,8 +108,8 @@ struct rtl2832_sdr_frame_buf {
 };
 
 struct rtl2832_sdr_dev {
-#define POWER_ON           (1 << 1)
-#define URB_BUF            (1 << 2)
+#define POWER_ON           0  /* BIT(0) */
+#define URB_BUF            1  /* BIT(1) */
 	unsigned long flags;
 
 	struct platform_device *pdev;
@@ -351,7 +351,7 @@ static int rtl2832_sdr_free_stream_bufs(struct rtl2832_sdr_dev *dev)
 {
 	struct platform_device *pdev = dev->pdev;
 
-	if (dev->flags & USB_STATE_URB_BUF) {
+	if (test_bit(URB_BUF, &dev->flags)) {
 		while (dev->buf_num) {
 			dev->buf_num--;
 			dev_dbg(&pdev->dev, "free buf=%d\n", dev->buf_num);
@@ -360,7 +360,7 @@ static int rtl2832_sdr_free_stream_bufs(struct rtl2832_sdr_dev *dev)
 					  dev->dma_addr[dev->buf_num]);
 		}
 	}
-	dev->flags &= ~USB_STATE_URB_BUF;
+	clear_bit(URB_BUF, &dev->flags);
 
 	return 0;
 }
@@ -389,7 +389,7 @@ static int rtl2832_sdr_alloc_stream_bufs(struct rtl2832_sdr_dev *dev)
 		dev_dbg(&pdev->dev, "alloc buf=%d %p (dma %llu)\n",
 			dev->buf_num, dev->buf_list[dev->buf_num],
 			(long long)dev->dma_addr[dev->buf_num]);
-		dev->flags |= USB_STATE_URB_BUF;
+		set_bit(URB_BUF, &dev->flags);
 	}
 
 	return 0;