Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1271311 > unrolled thread

[PATCH 1/8] Documentation/spi/spidev_test.c: use one rx buffer

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 1/8] Documentation/spi/spidev_test.c: use one rx buffer Joshua Clayton <stillcompiling@gmail.com> - 2015-11-17 16:30 +0100
    Re: [PATCH 1/8] Documentation/spi/spidev_test.c: use one rx buffer Mark Brown <broonie@kernel.org> - 2015-11-17 18:50 +0100
      Re: [PATCH 1/8] Documentation/spi/spidev_test.c: use one rx buffer Joshua Clayton <stillcompiling@gmail.com> - 2015-11-17 20:00 +0100

#1271311 — [PATCH 1/8] Documentation/spi/spidev_test.c: use one rx buffer

default_rx and rx are needlessly different.
Use one buffer, local to transmit()

Signed-off-by: Joshua Clayton <stillcompiling@gmail.com>
---
 Documentation/spi/spidev_test.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)

diff --git a/Documentation/spi/spidev_test.c b/Documentation/spi/spidev_test.c
index 135b3f5..dfe8f47 100644
--- a/Documentation/spi/spidev_test.c
+++ b/Documentation/spi/spidev_test.c
@@ -46,7 +46,6 @@ uint8_t default_tx[] = {
 	0xF0, 0x0D,
 };
 
-uint8_t default_rx[ARRAY_SIZE(default_tx)] = {0, };
 char *input_tx;
 
 static void hex_dump(const void *src, size_t length, size_t line_size, char *prefix)
@@ -100,10 +99,10 @@ static int unescape(char *_dst, char *_src, size_t len)
 	return ret;
 }
 
-static void transfer(int fd, uint8_t const *tx, uint8_t const *rx, size_t len)
+static void transfer(int fd, uint8_t const *tx, size_t len)
 {
 	int ret;
-
+	uint8_t *rx = malloc(len);
 	struct spi_ioc_transfer tr = {
 		.tx_buf = (unsigned long)tx,
 		.rx_buf = (unsigned long)rx,
@@ -135,6 +134,7 @@ static void transfer(int fd, uint8_t const *tx, uint8_t const *rx, size_t len)
 	if (verbose)
 		hex_dump(tx, len, 32, "TX");
 	hex_dump(rx, len, 32, "RX");
+	free(rx);
 }
 
 static void print_usage(const char *prog)
@@ -254,7 +254,6 @@ int main(int argc, char *argv[])
 	int ret = 0;
 	int fd;
 	uint8_t *tx;
-	uint8_t *rx;
 	int size;
 
 	parse_opts(argc, argv);
@@ -303,13 +302,11 @@ int main(int argc, char *argv[])
 	if (input_tx) {
 		size = strlen(input_tx+1);
 		tx = malloc(size);
-		rx = malloc(size);
 		size = unescape((char *)tx, input_tx, size);
-		transfer(fd, tx, rx, size);
-		free(rx);
+		transfer(fd, tx, size);
 		free(tx);
 	} else {
-		transfer(fd, default_tx, default_rx, sizeof(default_tx));
+		transfer(fd, default_tx, sizeof(default_tx));
 	}
 
 	close(fd);
-- 
2.5.0

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]

#1271491

[Multipart message — attachments visible in raw view] — view raw

On Tue, Nov 17, 2015 at 07:24:21AM -0800, Joshua Clayton wrote:

> default_rx and rx are needlessly different.
> Use one buffer, local to transmit()

Why?  This isn't what I'd expect from black boxing the API, from a
userspace point of view the transfer is atomic and in an ideal world
we'd be able to do direct to/from memory transfers rather than memcpy()
into kernel space which means that userspace should assume the transfers
are going on simultaneously even if they don't currently.

[toc] | [prev] | [next] | [standalone]

#1271567

On Tuesday, November 17, 2015 05:41:56 PM Mark Brown wrote:
> On Tue, Nov 17, 2015 at 07:24:21AM -0800, Joshua Clayton wrote:
> 
> > default_rx and rx are needlessly different.
> > Use one buffer, local to transmit()
> 
> Why?  This isn't what I'd expect from black boxing the API, from a
> userspace point of view the transfer is atomic and in an ideal world
> we'd be able to do direct to/from memory transfers rather than memcpy()
> into kernel space which means that userspace should assume the transfers
> are going on simultaneously even if they don't currently.

The important thing here was to get rid of the default_rx buffer.
I just noticed that the output can be set up completely within the scope
of the transmit function, since the operands are global. But I would be
just as happy to set it up at the top level. I'll change this in V2

--
~Joshua Clayton
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web