On 14 Feb 2000, Eric W. Biederman wrote:
> [EMAIL PROTECTED] (Eric W. Biederman) writes:
>
> > Anyway here is a patch that fixes it on my system.
>
> Grr. That wasn't the whole story.
For the shmat() this was exactly the right patch, and the below
has nothing to do with it ;-)
> Running wolf3d (with a suid dosemu) on the console I'm:
>
> EMS: mmap()ing from 0xe0000 to 0x40203000
> MAPPING: map, cap= EMS ALIAS, target=0xe0000, size=4000, protect=7, source=0x40203000
> EMS: mmap() failed: Invalid argument
This is the old /proc/self/mem mapping problem, which seems to reappear
in 2.2.14 and would need a further kludge. The kernel is returning -EINVAL
though it gets the right arguments and _should_ do its work.
This was the reason Linus abandoned mapping support of /proc/self/mem
for 2.3.x and which leads to the new patches in DOSEMU.
Can you pleas test if 2.2.13 doesn't show the bug? If yes, I'ld simple
tell the detection stuff to use the mapfile driver for 2.2.14.
I won't investigate any new work in new /proc/self/mem kludges ;-)
>
> I haven't tracked this one down yet.
> Passing:
> -I 'mappingdriver mapfile'
> to dosemu allows me to play just fine.
sure, it doesn't use /proc/self/mem atall ;-)
> I don't notice any signficant speed
> impact on my 486....
also normal, your cpu isn't faster than your disk ;-)
On a PII you can see the difference at a first glance: screen resfresh
becomes jerky, sudden delays in reaction e.t.c. In short all what can
happen if you access disk instead of RAM.
> I managed to get through all of the levels in 6:40 on easy...
good to know, so the mapfile driver seems ok.
>
> Hans mapself.c needs some more work, or maybe just plain ditching ....
No, look above, its the kernel who does weird things. On 2.0.x the
mapself driver has no problems at all.
Hans
<[EMAIL PROTECTED]>
