Re: bug in task server startup code

[cascardo]

I don't think it's related to this bug you've found, since it would make
an address 0 invalid access. In mem_add_task (or something like that), a
piece of code like:

first_task->prev = task;
first_task = task;

is a buggy one, since first_task in initialized to NULL.
A simple correction would be:

if (first_task != NULL)
  first_task->prev = task;
first_task = task;

Since it's very simple, I think you won't care about not having a full
patch file neither my signature of FSF papers.

I've seen address 4 being invalid accessed by tasks running over L4 when
using Doug Leas' malloc. I thinks this occurred when mmap did not return
the mmaped address. If malloc is used in physmem, this would shed some
light, I expect.

signature.asc
Description: Digital signature