Skip to contentSkip to conversation

build as module and fix priority inversion problem

PATCH In review Mark Underwood 2 participants 2 messages Dec 5, 2005

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 build as module and fix priority inversion problem

Sorry, made a bit of a mess with the last e-mail :(. Trying again :)

This patch fix's the possible priority inversion that vitaly pointed out and allows the driver to be built as a module.

David, please can you test this patch. Can you also submit a patch which adds the module author,
description and license please, thanks. Kconfig | 2 +-

 spi.c   |   23 ++++++++++++++++++-----
 2 files changed, 19 insertions(+), 6 deletions(-)

___________________________________________________________
Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail http://uk.messenger.yahoo.com

Kconfig +1 -1
diff -uprN spi-org/Kconfig spi/Kconfig
@@ -22,7 +22,7 @@ config SPI_ARCH_HAS_SLAVE
 	default y if ARCH_PXA
 
 config SPI
-	bool "SPI support"
+	tristate "SPI support"
 	depends on SPI_ARCH_HAS_MASTER || SPI_ARCH_HAS_SLAVE
 	help
 	  The "Serial Peripheral Interface" is a low level synchronous
spi.c +18 -5
diff -uprN spi-org/spi.c spi/spi.c
@@ -232,6 +232,8 @@ EXPORT_SYMBOL_GPL(spi_new_device);
  * The board info passed can safely be __initdata ... but be careful of
  * any embedded pointers (platform_data, etc), they're copied as-is.
  */
+
+#ifdef CONFIG_SPI
 int __init
 spi_register_board_info(struct spi_board_info const *info, unsigned n)
 {
@@ -249,6 +251,7 @@ spi_register_board_info(struct spi_board
 	return 0;
 }
 EXPORT_SYMBOL_GPL(spi_register_board_info);
+#endif
 
 /* FIXME someone should add support for a __setup("spi", ...) that
  * creates board info from kernel command lines
@@ -505,6 +508,7 @@ int spi_write_then_read(struct spi_devic
 	int			status;
 	struct spi_message	message;
 	struct spi_transfer	x[2];
+	u8			*local_buf;
 
 	/* Use preallocated DMA-safe buffer.  We can't avoid copying here,
 	 * (as a pure convenience thing), but we can keep heap costs
@@ -513,14 +517,19 @@ int spi_write_then_read(struct spi_devic
 	if ((n_tx + n_rx) > SPI_BUFSIZ)
 		return -EINVAL;
 
-	down(&lock);
+	if (down_trylock(&lock))
+		/* Someone else is using the main buffer, kalloc a new one */
+		local_buf = kmalloc(SPI_BUFSIZ, GFP_ATOMIC);
+	else
+		local_buf = buf;
+
 	memset(x, 0, sizeof x);
 
-	memcpy(buf, txbuf, n_tx);
-	x[0].tx_buf = buf;
+	memcpy(local_buf, txbuf, n_tx);
+	x[0].tx_buf = local_buf;
 	x[0].len = n_tx;
 
-	x[1].rx_buf = buf + n_tx;
+	x[1].rx_buf = local_buf + n_tx;
 	x[1].len = n_rx;
 
 	/* do the i/o */
@@ -532,7 +541,11 @@ int spi_write_then_read(struct spi_devic
 		status = message.status;
 	}
 
-	up(&lock);
+	if (x[0].tx_buf == buf)
+		up(&lock);
+	else
+		kfree(local_buf);
+
 	return status;
 }
 EXPORT_SYMBOL_GPL(spi_write_then_read);
#1 lore raw
Quoted 3 lines · Mark Underwood

Sorry, made a bit of a mess with the last e-mail :(. Trying again :)

This patch fix's the possible priority inversion that vitaly pointed out

That part looks OK to me, though I don't have time to test it just. It's the change to "use kmalloc if we must".

I'd like to see you split that part out so it can be merged into the next MM kernel. (I don't know when that is, but I'd guess RSN.)

Quoted 1 line

and allows the driver to be built as a module.

That part won't work right. Remember, the spi_board_info setup must be safe to call from arch_init() code; that support needs to be statically linked. The full patch for this would split those parts into a separate file, and expose some internal hooks to the module. (Like maybe the struct boardinfo type, and the board_list of instances, and the lock protecting that list.)

- Dave