RTO estimation changes. Under some circumstances it would return a value
of 0, while the old Van Jacobson RTO code would return a minimum of 3.
This would result in 12 retransmissions, each 1 second apart.
This takes care of those instances, and ensures that t_rttmin is
used everywhere as a lower bound.
the very rare case of shared mappings that have amap's attached in a
reasonable way -- this is not currently causing any problems, but i
fixed it anyway. update the comment in this section of code and also
be smarter about avoiding needless calls to pmap_protect().
change pfil_add_hook to put output filters at the tail of the queue,
while continuing to place input filters at the head of the queue. update
the two users of these functions, and document these changes.
fixes PR#4593.
invalidation algorithm:
if (old mapping had PG_ASM set || pmap is active) {
TIBS(va);
if (also sync I-stream)
imb();
}
The check for "old mapping had PG_ASM" will get all kernel mappings (since
kernel mappings always have PG_ASM set).
This allows us to remove the bogus check for the kernel pmap in
active_pmap() - do so.
Use the new TLB invalidation macro whenever such action is needed.
as well, until some other changes are made). Nuke active_user_pmap(),
and change the places that used it to use active_pmap() instead (as well
as make some DIAGNOSTIC consistency checks).
the PTE was previously invalid, no TLB invalidation is necessary because:
(1) when a PTE is invalidated, its entry is flushed from the
TLB
(2) the PALcode won't install an invalid PTE into the TLB.
structure. We will continue to run on this context (which is the
global Lev1map at this point) right up until we switch to proc0's
context in switch_exit().
to "options FFS_EI". The superblock and inodes (without blk addr) are
byteswapped at disk read/write time, other metadatas are byteswapped
when used (as they are acceeded directly in the buffer cache).
This required the addition of a "um_flags" field to struct ufsmount.
ffs_bswap.c contains superblock and inode byteswap routines also used
by userland utilities.
- Maintain two copies of the number and size of physical memory segments.
One copy, mem_clusters[], contains _all_ of physical RAM, for crash dumps.
The other copy, phys_seg_list[], starts out with all of physical RAM (and
is used to initialize mem_clusters[]), but is adjusted to be the memory
actually managed by the VM system.
- Fix computation of physmem; when support for MACHINE_NEW_NONCONTIG was
added, physmem was accidentally changed to not take into account the memory
located before the kernel, or the memory taken up by the kernel itself.
- Make initialization of the message buffer a little less magic-looking.
- Maintain two copies of the number and size of physical memory segments.
One copy, mem_clusters[], contains _all_ of physical RAM, for crash dumps.
The other copy, phys_seg_list[], starts out with all of physical RAM (and
is used to initialize mem_clusters[]), but is adjusted to be the memory
actually managed by the VM system.
- Fix crash dumps with regard to multiple memory segments.
in the packet. This fixes a bug that was resulting in extra packets
in retransmissions (the second packet would be 12 bytes long,
reflecting the RFC1323 timestamp option size).
vn, with a 0 component. If the upper fs was a unionfs,
union_whiteout() would deref compnent to get a struct proc, and panic.
struct proc was only being passed to FIXUP, which never used it. It
turns out this happened a lot. I ripped most of the unneeded code
out, and left in the few places that really did need the proc handle.
ragge