kgdb and console output
caution: perhaps a bit offtopic?
Hi all:
I'm still having problems to debug a kernel with kgdb:
both kgdb and console output go to /dev/ttyS0 (no chance of changing that, we use the second serial port for other purposes). I get an initial break in gdb (development machine), and the kernel loads smoothly (outputs are perfect) till it reaches init, when the output goes mad, repeating everything (but the first char) a number of times. For example:
INIT: INIT: NIT: IT: T: : version 2.78 bootingersion 2.78 bootingrsion 2.78 bootingsion 2.78 bootingon 2.78 booting
and so on.
a.- Is that a gdb problem?
b.- Is it related with using the same serial port for both output and debugging?
c.- Maybe i should change the tools i use in the root filesystem (aka /bin/echo ...)?
I have tried not using colors or scape sequences to print, but a simple echo "echo this" produces the same problem.
Any ideas?
Many regards.
--
Ibon Gotxi Garcia <[email protected]>
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
On Fri, Mar 14, 2003 at 11:49:50AM +0000, Ibon Gotxi Garcia wrote:
[snip]
Quoted 10 lines
INIT: INIT: NIT: IT: T: : version 2.78 bootingersion 2.78 bootingrsion 2.78 bootingsion 2.78 bootingon 2.78 booting
and so on.
a.- Is that a gdb problem?
b.- Is it related with using the same serial port for both output and debugging?
c.- Maybe i should change the tools i use in the root filesystem (aka/bin/echo...)?
It's related to the 'console' support inside of the serial driver / kgdb stub. The best way, IMHO, around this is to use the 'kgdb_demux' script. I don't have the URL handy, but google knows.
--
Tom Rini
http://gate.crashing.org/~trini/
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
Quoted 11 lines · Tom Rini
On Fri, Mar 14, 2003 at 11:49:50AM +0000, Ibon Gotxi Garcia wrote:
[snip]
Quoted 3 lines
INIT: INIT: NIT: IT: T: : version 2.78 bootingersion 2.78 bootingrsion 2.78 bootingsion 2.78 bootingon 2.78 booting
It's related to the 'console' support inside of the serial driver / kgdb stub. The best way, IMHO, around this is to use the '
kgdb_demux' script. I don't have the URL handy, but google knows.
I have found it in
http://linux.junsun.net/porting-howto/src/kdmx-1.02.tar.gz
but I'm still unable of making it work. I seems fine when the kernel is loading, but it begins to do funny thing when we move to userland.
I think the hole point is character scape: GDB packets are $data#checksum, so when some output (no debug output, just normal one) is made to the console having '$', '#' or even '+' and '-', the kdmx script goes mental.
I'm thinking in two possible solutions:
1.- Wait till kgdb has LAN support (gdb packets over udp, instead of over the serial port).
2.- Patch the console driver (no idea how) to avoid printing '$' or '#'. Probably a terrible hack, but perhaps useful for development.
Any idea about this?
Cheers.
Ibon.
Quoted 5 lines
--
Tom Rini
http://gate.crashing.org/~trini/
--
Ibon Gotxi Garcia <[email protected]>
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
Quoted 13 lines · Ibon Gotxi Garcia
Quoted 11 lines · Tom Rini
On Fri, Mar 14, 2003 at 11:49:50AM +0000, Ibon Gotxi Garcia wrote:
[snip]
Quoted 3 lines
INIT: INIT: NIT: IT: T: : version 2.78 bootingersion 2.78 bootingrsion 2.78 bootingsion 2.78 bootingon 2.78 booting
It's related to the 'console' support inside of the serial driver / kgdb stub. The best way, IMHO, around this is to use the '
kgdb_demux' script. I don't have the URL handy, but google knows.
Quoted 5 lines
I have found it in
http://linux.junsun.net/porting-howto/src/kdmx-1.02.tar.gzbut I'm still unable of making it work. I seems fine when the kernel is loading, but it begins to do funny thing when we move to userland.
Is this 8xx or something else? For 8xx, currently console output ALWAYS goes to KGDB, if KGDB is enabled. When the next 2.4.21-pre release comes out, this behavior will change. I suspect 8260 behaves in a similar manner, but I don't have the HW to test it, so I did not make a similar change there.
--
Tom Rini
http://gate.crashing.org/~trini/
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/