| Age | Commit message (Collapse) | Author |
|---|
|
after c2k9
allows buffer cache to be extended and grow/shrink dynamically
tested by many, ok oga@, "why not just commit it" deraadt@
|
|
fixed, but now it is time for a little break from the chaos.
ok kettenis
|
|
whether removing holes or parts of them is allowed or not.
Only allow hole removal in uvmspace_free(), when tearing the vmspace down.
ok art@
|
|
cache locality and will pave the way for the new pmemrange allocator.
Based on hints from art@ and ariane@.
ok ariane@, deraadt@, oga@
|
|
This has has been tested very very thoroughly on all archs we have
excepting 88k and 68k. Please see cvs log for the individual commit
messages.
ok beck@, thib@
|
|
which is exactly what the macro does.
Macro's that are nothing more then:
#define FUNCTION(arg) function(arg)
are almost always pointless and should go away.
OK blambert@
Agreed by many.
|
|
specifically, if we free a RELEASED anon, then we will first of all
remove the page from the anon, free the anon, then get the next page
relative to the anon page, then call uvm_pagefree().
The problem is that while we zero out anon->an_page, we do not zero out
pg->uanon. Now, uvm_pagefree() if pg->uanon is not NULL zeroes out some
variables in the struct for us. One of the backed out commits added more
zeroing there which would have exacerbated this use after free under
heavy paging (which was where we saw bugs). Fix this by zeroing out
pg->uanon.
I have looked for other similar cases, but have not found any as of yet.
been in snaps a while, "please do commit that" deraadt@
|
|
big uvm backout mess.
ok miod@
|
|
More backouts in line with previous ones, this appears to bring us back to a
stable condition.
A machine forced to 64mb of ram cycled 10GB through swap with this diff
and is still running as I type this. Other tests by ariane@ and thib@
also seem to show that it's alright.
ok deraadt@, thib@, ariane@
|
|
We still have no idea why this stops the crashes. but it does.
a machine forced to 64mb of ram cycled 10GB through swap with this diff
and is still running as I type this. Other tests by ariane@ and thib@
also seem to show that it's alright.
ok deraadt@, thib@, ariane@
|
|
This is for the same reason as the earlier backouts, to avoid the bug
either added or exposed sometime around c2k9. This *should* be the last
one.
prompted by deraadt@
ok ariane@
|
|
allocator).
"i can't see any obvious problems" oga
|
|
separately).
a change at or just before the hackathon has either exposed or added a
very very nasty memory corruption bug that is giving us hell right now.
So in the interest of kernel stability these diffs are being backed out
until such a time as that corruption bug has been found and squashed,
then the ones that are proven good may slowly return.
a quick hitlist of the main commits this backs out:
mine:
uvm_objwire
the lock change in uvm_swap.c
using trees for uvm objects instead of the hash
removing the pgo_releasepg callback.
art@'s:
putting pmap_page_protect(VM_PROT_NONE) in uvm_pagedeactivate() since
all callers called that just prior anyway.
ok beck@, ariane@.
prompted by deraadt@.
|
|
three
commits:
1) The sysctl allowing bufcachepercent to be changed at boot time.
2) The change moving the buffer cache hash chains to a red-black tree
3) The dynamic buffer cache (Which depended on the earlier too).
ok on the backout from marco and todd
|
|
> extend uvm_page_physload to have the ability to add "device" pages to the
> system.
since it was overlayed over a system that we warned would go "in to be
tested, but may be pulled out". oga, you just made me spend 20 minutes
of time I should not have had to spend doing this.
|
|
utter failure of process.
|
|
|
|
|
|
Also de-inline functions.
This diff is intended to ease debugging. Once the current bug is fixed,
proper inlining will be redone.
ok: oga
|
|
to function properly in the fast-allocation path (should not have triggered).
ok: oga, deraadt
|
|
ass-backwards.
afaik this was found by the LLVM CLang static analyser.
ok ariane@ a couple of days ago.
|
|
system.
This is needed in the case where you need managed pages so you can
handle faulting and pmap_page_protect() on said pages when you manage
memory in such regions (i'm looking at you, graphics cards).
these pages are flagged PG_DEV, and shall never be on the freelists,
assert this. behaviour remains unchanged in the non-device case,
specifically for all archs currently in the tree we panic if called
after bootstrap.
ok art@, kettenis@, ariane@, beck@.
|
|
|
|
media just a wee bit too much.
|
|
just move that into uvm_pagedeactivate.
oga@ ok
|
|
shitz.
Code mostly from netbsd with some tweaks from me. The loaning case is
specifcically if 0ed out right now because we do not do loaning.
ok ariane@, art@.
|
|
This commit won't change the default behaviour of the system unless the
buffer cache size is increased with sysctl kern.bufcachepercent. By default
our buffer cache is 10% of memory, which with this commit is now treated
as a low water mark. If the buffer cache size is increased, the new size
is treated as a high water mark and the buffer cache is permitted to grow
to that percentage of memory.
If the page daemon is invoked, the page daemon will ask the buffer cache
to relenquish pages. if the buffer cache has more than the low water mark it
will relenquish pages allowing them to be consumed by uvm. after a short
period the buffer cache will attempt to re-grow back to the high water mark.
This permits the use of a large buffer cache without penalizing the available
memory for other purposes.
Above the low water mark the buffer cache remains entirely subservient to
the page daemon, so if uvm requires pages, the buffer cache will abandon
them.
ok art@ thib@ oga@
|
|
"that's retarded" art@.
|
|
says were debugging leftovers.
ok ariane@
|
|
one by tedu@. It doesn't do anything smart yet, it just uses
plain old disksort. we also keep the old method of queueing bufs
since some miods have crazy MD drivers that need some love.
ok beck@, art@
tested by many on many archs.
|
|
|
|
global lock, switch the uvm object pages to being kept in a per-object
RB_TREE. Right now this is approximately the same speed, but cleaner.
When biglock usage is reduced this will improve concurrency due to lock
contention..
ok beck@ art@. Thanks to jasper for the speed testing.
|
|
not encrypted.
|
|
OK guenther@ otto@
|
|
pgo_releasepg() hook and just free the page the "normal" way in the one
place we'll ever see PG_RELEASED and should care (uvm_page_unbusy,
called in aiodoned).
ok art@, beck@, thib@
|
|
to free ranges.
Classify memory based on region with associated use-counter (which is used
to construct a priority list of where to allocate memory).
Based on code from tedu@, help from many.
Ok art@
|
|
so we won't hit the kassert a little bit below it.
"obviously better than what is there now" beck@
|
|
similar to the aobj.c changes, this one does vnodes. Vnodes are more
complex because they actaully have to sync to backing store. So firstly,
convert sync to sleep instead of setting released.
Now, for backing store, in the PGO_FREE case, if we set PG_RELEASED
*before* an async io, (nothing else will see it, the page is busy), then
we can ignore the page after the io is done. We could do something
similar for PGO_DEACTIVATE too, but that is another change. On error we
just clear the released flag, nothing else sets it for uobj pages other
than aiodoned.
ok thib@, beck@, ariane@
|
|
ok thib@, weingart@.
|
|
it is also not part of an aobj.
Clear anon flags at pagefree: page is no longer part of an anon.
ok oga
|
|
Makes trace in ddb useful.
ok oga
|
|
uvm_km deals with kernel memory which is either part of one of the
kernel maps, or the main kernel object (a uao). If on km_pgremove we hit
a busy page, just sleep on it, if so there's some async io (and that is
unlikely). we can remove the check for uvm_km_alloc1() for a released page
since now we will never end up with a removed but released page in the kernel
map (due to the other chunk and the last diff).
ok ariane@. Diff survived several make builds, on amd64 and sparc64,
also forced paging with ariane's evil program.
|
|
Now, the PG_ RELEASED flag currently has two (maybe three) uses. The
valid one is for use with async io where we want to free the page after
we've paged it out. The other ones are "oh i'd like to free this, but
someone else is playing with it". It's simpler to just sleep on the
damned page instead and stop the fiddling.
First step does uao's: in uao_detach, sleep on the object and free it
when we're clean, instead of setting a flag so it's freed after. In
uao_flush, do the same. Change the interation over the object in flush
so that we don't have to add marker pages or other such voodoo to the
list when we sleep (netbsd did that when they had a similar diff), just
use the hash always. We can now change uao_releasepg() to just free the
page, and not bother with the KILLME stuff. When the other objects are
fixed this hook will vanish.
Much discussion with art@ over the idea, and ariane@ over this specific
diff. As mentioned, this one is based loosely on a similar idea in
netbsd.
Been in my tree for a while, survived many make builds, etc, and forcing
paging using ariane's evil program.
ok ariane@, beck@.
|
|
sleep on them (and otherwise ignore them) sleep on the pointer to the
{aiodoned,pagedaemon}_proc members, and nuke the two extra words.
"no objections" art@, ok beck@.
|
|
two cases of pool_get() + memset(0) -> pool_get(,,,PR_ZERO)
1.5 cases of global variables are already zeroed, so don't zero them.
ok ariane@, comments on stuff i'd missed from blambert@ and cnst@.
|
|
nothing uses this code yet, but might as well do it the right way.
"if you can't live without commiting this." miod@
|
|
ariane@ ok.
|
|
recursive in some cases (mostly involving swapping). A proper fix is in
the works, but this will unbreak kernels for now.
|
|
pages.
"go for it" miod@
|
|
needed.
"of course" art@.
|
