summaryrefslogtreecommitdiff
path: root/BUGS
blob: 1c58c1813d0f7b422b552d9e3c48da113f6778d6 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
This file is for bugs in Mach.  Bugs in the Hurd go into TODO.

Do not ever delete a bug from this list; new Mach releases need to be
checked to make sure fixed bugs stay fixed, and so this list must
never get truncated.  -mib

--Reported, verified fixed--

--Reported, claimed fixed--
MiG user stubs need to destroy reply port on any message transmission
error, not just a limited set.

--Reported--
Bug in vm_fault: when vm_fault on a shadow page:
First page from anonymous copy; returns unavaliable.  Then fault on
the actual page and block.  This block is interruptible, but a lock is
being held on the copy object (because it's FIRST_M).  If the sleep is
interrupted, the lock gets released, but unfortunately, it could also
be suspended, and that's a lose.

pager flush of wired pages can block forever uninterruptibly.

unwiring of pages should call PAGE_WAKEUP

permission arg of vm_wire is not passed to vm_fault, so that vm_fault
overeagerly wires the current protection, which might be more than the
permission arg of vm_wire.

When wiring, vm_fault uses current permission instead of max
permission; this will cause a failure of vm_wire if the current
permission is later promoted.

Patch for MiG dealing when sreplyport and mach_port_poly_t are used together.