Nut upsdev - Aug 2009 - Driver fix for Ippon BCP800 (06da/0003), blazer_usb.c

Hi,

Thanks to some older discussions on this list I found it quite easy to
eventually track down problem with using USB cable with some more of
Megatec/Q1 UPSes. The trick (which is done in megatec_usb, but not in
blazer_usb) is to read in the junk that somehow gets into read buffers
before sending any commands. (I inserted debug statements so it's easy
to see that the data read is really junk: in my case it's most often a
middle piece of previous query and sometimes is appended with some
fragments of /proc/self/environment or something).

I will appreciate if you apply the patch to fix it (although probably
it's better to have some such flush function defined in usb wrappers
for every needing driver to use):

Index: drivers/blazer_usb.c
==================================================================---
drivers/blazer_usb.c	(revision 1896)
+++ drivers/blazer_usb.c	(working copy)
@@ -94,6 +94,23 @@
 }


+static int phoenix_flush_io()
+{
+	static
+	char buf[SMALLBUF+1];
+	int i, ret;
+	upsdebugx(3, "phoenix_flush_io");
+	for(i = 0; i < 10; i++) {
+		memset(buf, 0, sizeof(buf));
+		/* ret = usb->get_interrupt(udev, (unsigned char *)buf, 8, 1000); */
+		ret = usb_interrupt_read(udev, 0x81, buf, SMALLBUF, 1000);
+		upsdebugx(4, "flush(%d,%d) <- '%s'", ret, strlen(buf),
buf);
+		if (ret < 1)
+			break;
+	}
+	return 0;
+}
+
 static int phoenix_command(const char *cmd, char *buf, size_t buflen)
 {
 	char	tmp[SMALLBUF];
@@ -103,6 +120,7 @@
 	memset(tmp, 0, sizeof(tmp));
 	snprintf(tmp, sizeof(tmp), "%s", cmd);

+	phoenix_flush_io();
 	for (i = 0; i < strlen(tmp); i += ret) {

 		/* Write data in 8-byte chunks */

After this patch the driver seems to work OK.

Cheers,
-- Danya

Hi Danya,

2009/8/2 Danya Alexeyevsky <me.dendik at gmail.com>
> Hi,
>
> Thanks to some older discussions on this list I found it quite easy to
> eventually track down problem with using USB cable with some more of
> Megatec/Q1 UPSes. The trick (which is done in megatec_usb, but not in
> blazer_usb) is to read in the junk that somehow gets into read buffers
> before sending any commands. (I inserted debug statements so it's easy
> to see that the data read is really junk: in my case it's most often a
> middle piece of previous query and sometimes is appended with some
> fragments of /proc/self/environment or something).
>
> I will appreciate if you apply the patch to fix it (although probably
> it's better to have some such flush function defined in usb wrappers
> for every needing driver to use):
>
> Index: drivers/blazer_usb.c
> ==================================================================> ---
drivers/blazer_usb.c        (revision 1896)
> +++ drivers/blazer_usb.c        (working copy)
> @@ -94,6 +94,23 @@
>  }
>
>
> +static int phoenix_flush_io()
> +{
> +       static
> +       char buf[SMALLBUF+1];
> +       int i, ret;
> +       upsdebugx(3, "phoenix_flush_io");
> +       for(i = 0; i < 10; i++) {
> +               memset(buf, 0, sizeof(buf));
> +               /* ret = usb->get_interrupt(udev, (unsigned char *)buf,
8,
> 1000); */
> +               ret = usb_interrupt_read(udev, 0x81, buf, SMALLBUF, 1000);
> +               upsdebugx(4, "flush(%d,%d) <- '%s'",
ret, strlen(buf),
> buf);
> +               if (ret < 1)
> +                       break;
> +       }
> +       return 0;
> +}
> +
>  static int phoenix_command(const char *cmd, char *buf, size_t buflen)
>  {
>        char    tmp[SMALLBUF];
> @@ -103,6 +120,7 @@
>        memset(tmp, 0, sizeof(tmp));
>        snprintf(tmp, sizeof(tmp), "%s", cmd);
>
> +       phoenix_flush_io();
>        for (i = 0; i < strlen(tmp); i += ret) {
>
>                /* Write data in 8-byte chunks */
>
> After this patch the driver seems to work OK.
>
I'll leave up to Arjen and Alexander (the maintainers of this driver) to
approve this patch.

I've meanwhile added it to our patch tracker on Alioth:
https://alioth.debian.org/tracker/index.php?func=detail&aid=311863&group_id=30602&atid=411544

thanks for your contribution,
Arnaud
-- 
Linux / Unix Expert R&D - Eaton - http://www.eaton.com/mgeops
Network UPS Tools (NUT) Project Leader - http://www.networkupstools.org/
Debian Developer - http://www.debian.org
Free Software Developer - http://arnaud.quette.free.fr/
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.alioth.debian.org/pipermail/nut-upsdev/attachments/20090803/439da355/attachment.htm>

Citeren Danya Alexeyevsky <me.dendik op gmail.com>:
> Thanks to some older discussions on this list I found it quite easy to
> eventually track down problem with using USB cable with some more of
> Megatec/Q1 UPSes. The trick (which is done in megatec_usb, but not in
> blazer_usb) is to read in the junk that somehow gets into read buffers
> before sending any commands. (I inserted debug statements so it's easy
> to see that the data read is really junk: in my case it's most often a
> middle piece of previous query and sometimes is appended with some
> fragments of /proc/self/environment or something).
Thanks for this report. I have slightly modified your patch, but it  
should work out the same.
> I will appreciate if you apply the patch to fix it (although probably
> it's better to have some such flush function defined in usb wrappers
> for every needing driver to use):
The fact that we need to flush the IO buffers, really means that the  
USB to serial implementation is broken, so I'd prefer to do this  
explicitly in the driver.

Best regards, Arjen
-- 
Please keep list traffic on the list