From 94cef36797600d11a50d09828fa80df8a73dfd1c Mon Sep 17 00:00:00 2001 From: "Michael I. Bushnell" Date: Thu, 18 Jul 1996 04:35:29 +0000 Subject: *** empty log message *** --- BUGS | 33 +++++++++++++++++++++++++++++++++ 1 file changed, 33 insertions(+) create mode 100644 BUGS (limited to 'BUGS') diff --git a/BUGS b/BUGS new file mode 100644 index 00000000..1c58c181 --- /dev/null +++ b/BUGS @@ -0,0 +1,33 @@ +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. -- cgit v1.2.3