sane-project-website/old-archive/1999-04/0094.html

<!-- received="Mon Apr  5 23:04:04 1999 PDT" -->
<!-- sent="Tue, 6 Apr 1999 08:03:27 +0200 (MET DST)" -->
<!-- name="=?ISO-8859-1?Q?Mikko_Ty=F6l=E4j=E4rvi?=" -->
<!-- email="mikko@securitydynamics.com" -->
<!-- subject="Bugs &amp; Fixes: FreeBSD, Snapscan, Sane 1.01-4" -->
<!-- id="" -->
<!-- inreplyto="" -->
<title>sane-devel: Bugs &amp; Fixes: FreeBSD, Snapscan, Sane 1.01-4</title>
<h1>Bugs &amp; Fixes: FreeBSD, Snapscan, Sane 1.01-4</h1>
<b>=?ISO-8859-1?Q?Mikko_Ty=F6l=E4j=E4rvi?=</b> (<a href="mailto:mikko@securitydynamics.com"><i>mikko@securitydynamics.com</i></a>)<br>
<i>Tue, 6 Apr 1999 08:03:27 +0200 (MET DST)</i>
<p>
<ul>
<li> <b>Messages sorted by:</b> <a href="date.html#94">[ date ]</a><a href="index.html#94">[ thread ]</a><a href="subject.html#94">[ subject ]</a><a href="author.html#94">[ author ]</a>
<!-- next="start" -->
<li> <b>Next message:</b> <a href="0095.html">killian@lucent.com: "Epson GT-7000"</a>
<li> <b>Previous message:</b> <a href="0093.html">De Clarke: "Bernd's patch for pre 1.01-4 for microtek2"</a>
<!-- nextthread="start" -->
<!-- reply="end" -->
</ul>
<!-- body="start" -->
Hi!<br>
<p>
I finally got around to changing SCSI-cards and testing if my scanner<br>
still worked after moving to FreeBSD 3.1.  Of course it didn't...<br>
But after a few sessions with truss and gdb, I can now scan again.<br>
<p>
For the records:<br>
<p>
 FreeBSD 3.1, Adaptec 1542B, AGFA Snapscan 310, sane-pre1.01-4<br>
<p>
<p>
	Regards,<br>
	/Mikko<br>
<p>
<p>
Problem 1:<br>
<p>
The snapscan backend does not work with FreeBSD 3.1 (the CAM SCSI<br>
stuff).  The reason is that CAM uses a faked fd that has nothing in<br>
common with a unix file descriptor.  Consequently doing a dup2() on<br>
it will not work too well.<br>
<p>
As far as I can see there is no reason to move the scanner to stdin in<br>
the child process, and just removing the operation indeed made<br>
scanning work.<br>
<p>
NOTE: It was possible to scan with "scanimage", but that is just<br>
coincidence (the CAM device happens to be at index 0 == STDIN_FILENO)<br>
<p>
A quick grep seems to indicate that no other backend suffers<br>
from this quirk.<br>
<p>
<p>
--- backend/snapscan.c.org	Mon Apr  5 14:26:47 1999<br>
+++ backend/snapscan.c	Mon Apr  5 14:35:11 1999<br>
@@ -2843,10 +2843,8 @@<br>
          and stdout the pipe */<br>
 	  signal (SIGTERM, handler);<br>
 	  reader_pss = pss;<br>
-	  dup2 (pss-&gt;fd, STDIN_FILENO);<br>
 	  dup2 (pss-&gt;rpipe[1], STDOUT_FILENO);<br>
 	  close (pss-&gt;rpipe[0]);<br>
-	  pss-&gt;fd = STDIN_FILENO;<br>
 	  reader (pss);<br>
 	  /* regular exit will cause a SIGPIPE */<br>
 	  DBG (DL_MINOR_INFO, "Reader process terminating.\n");<br>
<p>
<p>
<p>
Problem 2:<br>
<p>
The CAM code will overflow when using "xscanimage".  Every time a<br>
device is opened, a new slot in a bounded array is allocated. It is<br>
never re-used, even when the device is closed.  A few previews and<br>
scans in the same "xscanimage" process is all it takes to use up all<br>
the slots, and start writing outside allocated memory.<br>
<p>
There is no need to copy the data from cam_open_device(), it is<br>
already allocated, and will be freed by cam_close_device().  In fact,<br>
copying the data will only result in a memory leak.<br>
<p>
<p>
Minor problem 2.1:<br>
<p>
The cam_open_device() function requires access to /dev/xpt0, rather<br>
than (for example) /dev/pass4, where the scanner is, as well as the<br>
device being called "pass&lt;N&gt;".<br>
<p>
Using cam_open_pass() instead goes directly for the pass device,<br>
saving users the headache of figuring out why it will only work when<br>
running as "root", even though anyone can read &amp; write /dev/pass4, or<br>
why it will not work with a symlink.<br>
<p>
<p>
--- sanei/sanei_scsi.c.org	Mon Apr  5 14:26:35 1999<br>
+++ sanei/sanei_scsi.c	Mon Apr  5 14:58:17 1999<br>
@@ -197,7 +197,7 @@<br>
 <br>
 #if USE == FREEBSD_CAM_INTERFACE<br>
 # define CAM_MAXDEVS	128<br>
-struct cam_device **cam_devices = NULL;<br>
+struct cam_device *cam_devices[CAM_MAXDEVS] = { NULL };<br>
 #endif<br>
 <br>
 static struct<br>
@@ -817,23 +817,22 @@<br>
 		   just '{' unfortunately, this only works if all of<br>
 		   the '{' are that way. */<br>
 <br>
-      static int		unit = -1;<br>
-      struct cam_device*	curdev = NULL;<br>
+      struct cam_device *curdev;<br>
 <br>
-      fake_fd = -1;<br>
+      fake_fd = 1;<br>
       fd = -1;<br>
 <br>
-      if(cam_devices == NULL) {<br>
-	 size_t ary_sz = sizeof(struct cam_device*) * (CAM_MAXDEVS + 1);<br>
-	 cam_devices = malloc(ary_sz);<br>
-	 bzero(cam_devices, ary_sz);<br>
-	 unit = -1;	/* redundant, I know. */<br>
-      }<br>
+      if((curdev = cam_open_pass(dev, O_RDWR, NULL)) != NULL) {<br>
+	for (fd = 0; fd &lt; CAM_MAXDEVS &amp;&amp; cam_devices[fd] != NULL; fd++)<br>
+	  ;<br>
 <br>
-      if((curdev = cam_open_device(dev, O_RDWR)) != NULL) {<br>
-	cam_devices[++unit] = malloc(sizeof(struct cam_device));<br>
-	bcopy(curdev, cam_devices[unit], sizeof(struct cam_device));<br>
-	fd = unit;<br>
+	if (fd == CAM_MAXDEVS)<br>
+	  {<br>
+	    DBG(1, "sanei_scsi_open: too many CAM devices (%d)\n", fd);<br>
+	    cam_close_device(curdev);<br>
+	    return SANE_STATUS_INVAL;<br>
+	  }<br>
+	cam_devices[fd] = curdev;<br>
       }<br>
       else {<br>
 	 DBG(1, "sanei_scsi_open: device %s not found\n",<br>
@@ -969,7 +968,8 @@<br>
     close (fd);<br>
 <br>
 #if USE == FREEBSD_CAM_INTERFACE<br>
-   cam_close_device(cam_devices[fd]);<br>
+  cam_close_device(cam_devices[fd]);<br>
+  cam_devices[fd] = NULL;<br>
 #elif USE == DOMAINOS_INTERFACE<br>
   {<br>
     static int index;<br>
@@ -1689,7 +1689,7 @@<br>
    char*		data_buf;<br>
    size_t		data_len;<br>
    <br>
-   if (fd &lt; 0) {<br>
+   if (fd &lt; 0 || fd &gt; CAM_MAXDEVS || cam_devices[fd] != NULL) {<br>
       fprintf(stderr, "attempt to reference invalid unit %d\n", fd);<br>
       return SANE_STATUS_INVAL;<br>
    }<br>
<p>
<p>
 Mikko Tyo"la"ja"rvi________________________________mikko@securitydynamics.com<br>
 SecurityDynamics<br>
<p>
<p>
<pre>
--
Source code, list archive, and docs: <a href="http://www.mostang.com/sane/">http://www.mostang.com/sane/</a>
To unsubscribe: echo unsubscribe sane-devel | mail <a href="mailto:majordomo@mostang.com">majordomo@mostang.com</a>
</pre>
<!-- body="end" -->
<p>
<ul>
<!-- next="start" -->
<li> <b>Next message:</b> <a href="0095.html">killian@lucent.com: "Epson GT-7000"</a>
<li> <b>Previous message:</b> <a href="0093.html">De Clarke: "Bernd's patch for pre 1.01-4 for microtek2"</a>
<!-- nextthread="start" -->
<!-- reply="end" -->
</ul>