Hello Michael , Hello Hackers,
i have tried the patch also. It seems, that l4linux-2.0.21 is more stable than before. Also the module package seems to work (dummy.o ne.o and related). That means in my case that these modules are autoloaded (ping works). But i havenĀ“t a real network.
Occasianally the system does not reached the login, unfortunatly i have no copy of the given message (probably rcv fault, during /sbin/init).
Other times the reboot command causes a seg fault (null pointer derefenrence) after the message "Send all processes kill sig" (This is reported earlier).
(I have (re)configured in serial support but disabled com2 in the bios (i have no remote debugger). So it makes no difference in using -nokdb or not.)
Regards Harald Fritzsche
Michael Hohmuth wrote:
roux@ecoledoc.lip6.fr (Cedric ROUX) writes:
Then, another time, the boot process blocks with this time the message : ipc.cc:169 : failed assertion `! in_sender_list()'
Attached below are some Fiasco patches which should help.
This problem was a symptom of a familiar problem: Code which disables interrupts to protect a critical region trapped into a page fault which subsequently enabled interrupts again. We now protect against this problem by pre-installing all pages required for the operation in a safe way before the actual critical region starts.
Ugly synchronization problems like this have been one of the main reasons to try to get rid of disabling interrupts altogether. However, the occasional "cli" instruction is still lurking in the code, waiting to be replaced with better mechanisms (like non-blocking synchronization).
The patch also includes a fix which enhances compatibility with Jochen's L4 so that L4Linux can handle segmentation faults better.
Michael
hohmuth@innocent.com, hohmuth@inf.tu-dresden.de http://home.pages.de/~hohmuth/
Name: patch3.diff
patch3.diff Type: unspecified type (application/octet-stream) Encoding: 7bit