14287 followup - console

Part of the original submission that resulted in 14287 was to catch all output to the console/vga. 14287, as committed, only captures a portion of the console output. The attached patch allows capture of the rest of the output.

-b

------------------------------------------------------------------------
Ben Thomas                                         Virtual Iron Software
[email protected]                            Tower 1, Floor 2
978-849-1214                                       900 Chelmsford Street
                                                   Lowell, MA 01851

diff -r f136ed65b65a xen/drivers/char/console.c
@@ -328,8 +328,12 @@ static long guest_console_write(XEN_GUES
 
         sercon_puts(kbuf);
 
-        for ( kptr = kbuf; *kptr != '\0'; kptr++ )
+        for ( kptr = kbuf; *kptr != '\0'; kptr++ ) {
             vga_putchar(*kptr);
+            putchar_console_ring(*kptr);
+        }
+
+        send_guest_global_virq(dom0, VIRQ_CON_RING);
 
         guest_handle_add_offset(buffer, kcount);
         count -= kcount;

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

The console ring is only for Xen output (and also, as it happens, HVM domU bootstrap output since bytes output via port 0xe9 also get printk'ed). The only other useful output would be from dom0, and you can already obtain that via syslog. By not capturing dom0 output in the console ring we prevent interactive sessions on the dom0 console from obliterating useful Xen debug output which we may want to gather later (if we're not running an automatic gatherer daemon).

-- Keir

Hi,

I understand what you're saying; it was one of the considerations when first putting the patch together some time ago. The console ring was used for two reasons: the primary reason was to get a more WYSIWYG output. Two input sources are inherently capable of getting slightly out of sync (or worse). The second reason is that syslog can be expensive. In the end, the WYSIWYG and cost resulted in the console ring approach.

Would this be more acceptable if it were selectable ? That is, come up with a libxc/sysctl mechanism (which should be pretty easy) to control the gating of character data to the console ring fom guest_console_write ?

The capability is pretty useful in remoting the vga output to another node (and logging it). It would be nice to find a way to preserve current behavior (as you note), and also allow the ability to continue to remote it.

Thanks,
-b

------------------------------------------------------------------------
Ben Thomas                                         Virtual Iron Software
[email protected]                            Tower 1, Floor 2
978-849-1214                                       900 Chelmsford Street
                                                   Lowell, MA 01851

You could add a Xen boot parameter.

-- Keir

Sounds like a plan...

How about this ?

-b

------------------------------------------------------------------------
Ben Thomas                                         Virtual Iron Software
[email protected]                            Tower 1, Floor 2
978-849-1214                                       900 Chelmsford Street
                                                   Lowell, MA 01851

Add boot option to allow console I/O to be placed into ring buffer

This option allows console I/O to be placed into the hypervisor console ring buffer. When enabled, all output to the console device will be reflected in the ring buffer. When disabled (the default), only hypervisor output is available from the ring buffer.

diff -r f183e0a4d0c9 docs/src/user.tex
@@ -3179,6 +3179,11 @@ editing \path{grub.conf}.
   one of 80x25, 80x28, 80x30, 80x34, 80x43, 80x50, 80x60.
   \item[ keep ] Keep the VGA console even after domain 0 boots.
   \end{description}
+\item [ console\_to\_ring ] Place guest console output into the
+  hypervisor console ring buffer. This is disabled by default.
+  When enabled, both hypervisor output and guest console output
+  is available from the ring buffer. This can be useful for logging
+  and/or remote presentation of console data.
 \item [ sync\_console ] Force synchronous console output. This is
   useful if you system fails unexpectedly before it has sent all
   available output to the console. In most cases Xen will

diff -r f183e0a4d0c9 xen/drivers/char/console.c
@@ -48,6 +48,10 @@ static int opt_sync_console;
 static int opt_sync_console;
 boolean_param("sync_console", opt_sync_console);
 
+/* console_to_ring: send guest (incl dom 0) console data to console ring */
+static int opt_console_to_ring;
+boolean_param("console_to_ring", opt_console_to_ring);
+
 #define CONRING_SIZE 16384
 #define CONRING_IDX_MASK(i) ((i)&(CONRING_SIZE-1))
 static char conring[CONRING_SIZE];

@@ -328,8 +332,14 @@ static long guest_console_write(XEN_GUES
 
         sercon_puts(kbuf);
 
-        for ( kptr = kbuf; *kptr != '\0'; kptr++ )
+        for ( kptr = kbuf; *kptr != '\0'; kptr++ ) {
             vga_putchar(*kptr);
+            if (opt_console_to_ring)
+                putchar_console_ring(*kptr);
+        }
+
+        if (opt_console_to_ring)
+            send_guest_global_virq(dom0, VIRQ_CON_RING);
 
         guest_handle_add_offset(buffer, kcount);
         count -= kcount;

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel