[PATCH] Faster Dediprog writes

[Carl-Daniel Hailfinger]

Implementing bulk write for max speed on the Dediprog will require some
traces of the SPI bus.

Add chunked write ability to the Dediprog SF100 driver.
The chunk size is guessed, you can try to raise it up to 256, but please
check that write works without corruption. The chunk size may also
depend on the firmware version. 12 should be a safe bet, and if
everything else fails, reduce the chunk size to 1 (old behaviour).

Please note that the write speedup only applies to chips which have SPI
page write (i.e. chips using spi_chip_write_256).

Tests appreciated.

Signed-off-by: Carl-Daniel Hailfinger

Index: flashrom-dediprog_write_chunked/dediprog.c
===================================================================
--- flashrom-dediprog_write_chunked/dediprog.c	(Revision 1234)
+++ flashrom-dediprog_write_chunked/dediprog.c	(Arbeitskopie)
@@ -238,6 +238,12 @@
    return 0;
 }

+int dediprog_spi_write_256(struct flashchip *flash, uint8_t *buf, int start, int len)
+{
+	/* No idea about the real limit. Maybe 12, maybe more, maybe less. */
+	return spi_write_chunked(flash, buf, start, len, 12);
+}
+
 int dediprog_spi_send_command(unsigned int writecnt, unsigned int readcnt,
    		const unsigned char *writearr, unsigned char *readarr)
 {
Index: flashrom-dediprog_write_chunked/spi.c
===================================================================
--- flashrom-dediprog_write_chunked/spi.c	(Revision 1234)
+++ flashrom-dediprog_write_chunked/spi.c	(Arbeitskopie)
@@ -124,7 +124,7 @@
    	.command = dediprog_spi_send_command,
    	.multicommand = default_spi_send_multicommand,
    	.read = dediprog_spi_read,
-		.write_256 = spi_chip_write_1,
+		.write_256 = dediprog_spi_write_256,
    },
 #endif

Index: flashrom-dediprog_write_chunked/programmer.h
===================================================================
--- flashrom-dediprog_write_chunked/programmer.h	(Revision 1234)
+++ flashrom-dediprog_write_chunked/programmer.h	(Arbeitskopie)
@@ -472,6 +472,7 @@
 int dediprog_shutdown(void);
 int dediprog_spi_send_command(unsigned int writecnt, unsigned int readcnt, const unsigned char *writearr, unsigned char *readarr);
 int dediprog_spi_read(struct flashchip *flash, uint8_t *buf, int start, int len);
+int dediprog_spi_write_256(struct flashchip *flash, uint8_t *buf, int start, int len);

/* flashrom.c */
 struct decode_sizes {

[Richard Smith]

On Wed, Nov 17, 2010 at 11:53 PM, Carl-Daniel Hailfinger
 wrote:
> Please note that the write speedup only applies to chips which have SPI
> page write (i.e. chips using spi_chip_write_256).
>
> Tests appreciated.
>
> Signed-off-by: Carl-Daniel Hailfinger

I had to tweak the   if (writecnt > 5) {} clause to allow for the
larger size but other than that this Works For Me. It drops the
program time on my 128k part to 30 seconds.  Thats quite a bit better
than the 20 minutes it used to take.

I also verified that chunk sizes larger than 12 fail but I didn't try
any values between 1 and 12.

[Carl-Daniel Hailfinger]

On 18.11.2010 16:37, Richard Smith wrote:
> On Wed, Nov 17, 2010 at 11:53 PM, Carl-Daniel Hailfinger
>  wrote:
> > Please note that the write speedup only applies to chips which have SPI
> > page write (i.e. chips using spi_chip_write_256).
> >
> > Tests appreciated.
> >
> > Signed-off-by: Carl-Daniel Hailfinger
>
> I had to tweak the   if (writecnt > 5) {} clause to allow for the
> larger size but other than that this Works For Me. It drops the
> program time on my 128k part to 30 seconds.  Thats quite a bit better
> than the 20 minutes it used to take.
>
> I also verified that chunk sizes larger than 12 fail but I didn't try
> any values between 1 and 12.

Thanks for testing! New patch follows. If I get an ack, I'll commit.

Add chunked write ability to the Dediprog SF100 driver.

Please note that the write speedup only applies to chips which have SPI
page write (i.e. chips using spi_chip_write_256).

This is a quick fix for write speed until I get around to implementing
full bulk SPI write support.

Tests appreciated.

Signed-off-by: Carl-Daniel Hailfinger

Index: flashrom-dediprog_write_chunked/dediprog.c
===================================================================
--- flashrom-dediprog_write_chunked/dediprog.c	(Revision 1234)
+++ flashrom-dediprog_write_chunked/dediprog.c	(Arbeitskopie)
@@ -238,6 +238,12 @@
    return 0;
 }

+int dediprog_spi_write_256(struct flashchip *flash, uint8_t *buf, int start, int len)
+{
+	/* No idea about the real limit. Maybe 12, maybe more, maybe less. */
+	return spi_write_chunked(flash, buf, start, len, 12);
+}
+
 int dediprog_spi_send_command(unsigned int writecnt, unsigned int readcnt,
    		const unsigned char *writearr, unsigned char *readarr)
 {
@@ -245,7 +251,7 @@

msg_pspew("%s, writecnt=%i, readcnt=%i\n", __func__, writecnt, readcnt);
    /* Paranoid, but I don't want to be blamed if anything explodes. */
-	if (writecnt > 5) {
+	if (writecnt > 16) {
    	msg_perr("Untested writecnt=%i, aborting.\n", writecnt);
    	return 1;
    }
Index: flashrom-dediprog_write_chunked/spi.c
===================================================================
--- flashrom-dediprog_write_chunked/spi.c	(Revision 1234)
+++ flashrom-dediprog_write_chunked/spi.c	(Arbeitskopie)
@@ -124,7 +124,7 @@
    	.command = dediprog_spi_send_command,
    	.multicommand = default_spi_send_multicommand,
    	.read = dediprog_spi_read,
-		.write_256 = spi_chip_write_1,
+		.write_256 = dediprog_spi_write_256,
    },
 #endif

Index: flashrom-dediprog_write_chunked/programmer.h
===================================================================
--- flashrom-dediprog_write_chunked/programmer.h	(Revision 1234)
+++ flashrom-dediprog_write_chunked/programmer.h	(Arbeitskopie)
@@ -472,6 +472,7 @@
 int dediprog_shutdown(void);
 int dediprog_spi_send_command(unsigned int writecnt, unsigned int readcnt, const unsigned char *writearr, unsigned char *readarr);
 int dediprog_spi_read(struct flashchip *flash, uint8_t *buf, int start, int len);
+int dediprog_spi_write_256(struct flashchip *flash, uint8_t *buf, int start, int len);

/* flashrom.c */
 struct decode_sizes {

[Richard A. Smith]

On 11/22/2010 06:54 PM, Carl-Daniel Hailfinger wrote:
> On 18.11.2010 16:37, Richard Smith wrote:
> > On Wed, Nov 17, 2010 at 11:53 PM, Carl-Daniel Hailfinger
> >   wrote:
> > > Please note that the write speedup only applies to chips which have SPI
> > > page write (i.e. chips using spi_chip_write_256).
> > >
> > > Tests appreciated.
> > >
> > > Signed-off-by: Carl-Daniel Hailfinger
> >
> > I had to tweak the   if (writecnt>  5) {} clause to allow for the
> > larger size but other than that this Works For Me. It drops the
> > program time on my 128k part to 30 seconds.  Thats quite a bit better
> > than the 20 minutes it used to take.
> >
> > I also verified that chunk sizes larger than 12 fail but I didn't try
> > any values between 1 and 12.
>
> Thanks for testing! New patch follows. If I get an ack, I'll commit.
>
> Add chunked write ability to the Dediprog SF100 driver.
>
> Please note that the write speedup only applies to chips which have SPI
> page write (i.e. chips using spi_chip_write_256).
>
> This is a quick fix for write speed until I get around to implementing
> full bulk SPI write support.
>
> Tests appreciated.

Acked-by: Richard A. Smith

[Mathias Krause]

On 23.11.2010 00:54, Carl-Daniel Hailfinger wrote:
> On 18.11.2010 16:37, Richard Smith wrote:
> > On Wed, Nov 17, 2010 at 11:53 PM, Carl-Daniel Hailfinger
> >  wrote:
> > > Please note that the write speedup only applies to chips which have SPI
> > > page write (i.e. chips using spi_chip_write_256).
> > >
> > > Tests appreciated.
> > >
> > > Signed-off-by: Carl-Daniel Hailfinger
> >
> > I had to tweak the   if (writecnt > 5) {} clause to allow for the
> > larger size but other than that this Works For Me. It drops the
> > program time on my 128k part to 30 seconds.  Thats quite a bit better
> > than the 20 minutes it used to take.
> >
> > I also verified that chunk sizes larger than 12 fail but I didn't try
> > any values between 1 and 12.
>
> Thanks for testing! New patch follows. If I get an ack, I'll commit.
>
> Add chunked write ability to the Dediprog SF100 driver.
>
> Please note that the write speedup only applies to chips which have SPI
> page write (i.e. chips using spi_chip_write_256).
>
> This is a quick fix for write speed until I get around to implementing
> full bulk SPI write support.
>
> Tests appreciated.
>
> Signed-off-by: Carl-Daniel Hailfinger

Acked-by: Mathias Krause

[Carl-Daniel Hailfinger]

On 23.11.2010 08:30, Mathias Krause wrote:
> On 23.11.2010 00:54, Carl-Daniel Hailfinger wrote:
> > On 18.11.2010 16:37, Richard Smith wrote:
> > > On Wed, Nov 17, 2010 at 11:53 PM, Carl-Daniel Hailfinger
> > >  wrote:
> > > > Please note that the write speedup only applies to chips which have SPI
> > > > page write (i.e. chips using spi_chip_write_256).
> > > >
> > > > Tests appreciated.
> > > >
> > > > Signed-off-by: Carl-Daniel Hailfinger
> > >
> > > I had to tweak the   if (writecnt > 5) {} clause to allow for the
> > > larger size but other than that this Works For Me. It drops the
> > > program time on my 128k part to 30 seconds.  Thats quite a bit better
> > > than the 20 minutes it used to take.
> > >
> > > I also verified that chunk sizes larger than 12 fail but I didn't try
> > > any values between 1 and 12.
> >
> > Thanks for testing! New patch follows. If I get an ack, I'll commit.
> >
> > Add chunked write ability to the Dediprog SF100 driver.
> >
> > Please note that the write speedup only applies to chips which have SPI
> > page write (i.e. chips using spi_chip_write_256).
> >
> > This is a quick fix for write speed until I get around to implementing
> > full bulk SPI write support.
> >
> > Tests appreciated.
> >
> > Signed-off-by: Carl-Daniel Hailfinger
>
> Acked-by: Mathias Krause

Thanks, committed in r1235.

Regards,
Carl-Daniel