We're moving towards a world in which purging decisions are less rigidly
enforced at a single-hugepage level. In that world, it makes sense to keep
around some hpdatas which are not completely purged, in which case we'll need to
track them.
Really, this isn't a functional change, just a naming change. We start thinking
of pageslabs as being always in the psset. What we used to think of as removal
is now thought of as being in the psset, but in the process of being updated
(and therefore, unavalable for serving new allocations).
This is in preparation of subsequent changes to support deferred purging;
allocations will still be in the psset for the purposes of choosing when to
purge, but not for purposes of allocation/deallocation.
This is really only useful for human consumption. Correspondingly, emit it only
in the human-readable stats, and let everybody else compute from the hugepage
size and nactive.
Previously, we would purge a hugepage only when it's completely empty. With
this change, we can purge even when only partially empty. Although the
heuristic here is still fairly primitive, this infrastructure can scale to
become more advanced.
This saves us a cache miss when lookup up the arena bin offset in a remote
arena during tcache flush. All arenas share the base offset, and so we don't
need to look it up repeatedly for each arena. Secondarily, it shaves 288 bytes
off the arena on, e.g., x86-64.
The items we pick to flush matter a lot, but the order in which they get flushed
doesn't; just use forward scans. This simplifies the accessing code, both in
terms of the C and the generated assembly (i.e. this speeds up the flush
pathways).
By carefully force-inlining the division constants and the operation sum count,
we can eliminate redundant operations in the arena-level dalloc function. Do
so.
This frontloads more of the miss latency. It also moves it to a pathway where
we have not yet acquired any locks, so that it should (hopefully) reduce hold
times.
In practice, many rtree_leaf_elm accesses are cache misses. By restructuring,
we can make it more likely that these misses occur without blocking us from
starting later lookups, taking more of those misses in parallel.
qemu does not support this, yet [1], and you can get very tricky assert
if you will run program with jemalloc in use under qemu:
<jemalloc>: ../contrib/jemalloc/src/extent.c:1195: Failed assertion: "p[i] == 0"
[1]: https://patchwork.kernel.org/patch/10576637/
Here is a simple example that shows the problem [2]:
// Gist to check possible issues with MADV_DONTNEED
// For example it does not supported by qemu user
// There is a patch for this [1], but it hasn't been applied.
// [1]: https://lists.gnu.org/archive/html/qemu-devel/2018-08/msg05422.html
#include <sys/mman.h>
#include <stdio.h>
#include <stddef.h>
#include <assert.h>
#include <string.h>
int main(int argc, char **argv)
{
void *addr = mmap(NULL, 1<<16, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
if (addr == MAP_FAILED) {
perror("mmap");
return 1;
}
memset(addr, 'A', 1<<16);
if (!madvise(addr, 1<<16, MADV_DONTNEED)) {
puts("MADV_DONTNEED does not return error. Check memory.");
for (int i = 0; i < 1<<16; ++i) {
assert(((unsigned char *)addr)[i] == 0);
}
} else {
perror("madvise");
}
if (munmap(addr, 1<<16)) {
perror("munmap");
return 1;
}
return 0;
}
### unpatched qemu
$ qemu-x86_64-static /tmp/test-MADV_DONTNEED
MADV_DONTNEED does not return error. Check memory.
test-MADV_DONTNEED: /tmp/test-MADV_DONTNEED.c:19: main: Assertion `((unsigned char *)addr)[i] == 0' failed.
qemu: uncaught target signal 6 (Aborted) - core dumped
Aborted (core dumped)
### patched qemu (by returning ENOSYS error)
$ qemu-x86_64 /tmp/test-MADV_DONTNEED
madvise: Success
### patch for qemu to return ENOSYS
diff --git a/linux-user/syscall.c b/linux-user/syscall.c
index 897d20c076..5540792e0e 100644
--- a/linux-user/syscall.c
+++ b/linux-user/syscall.c
@@ -11775,7 +11775,7 @@ static abi_long do_syscall1(void *cpu_env, int num, abi_long arg1,
turns private file-backed mappings into anonymous mappings.
This will break MADV_DONTNEED.
This is a hint, so ignoring and returning success is ok. */
- return 0;
+ return ENOSYS;
#endif
#ifdef TARGET_NR_fcntl64
case TARGET_NR_fcntl64:
[2]: https://gist.github.com/azat/12ba2c825b710653ece34dba7f926ece
v2:
- review fixes
- add opt_dont_trust_madvise
v3:
- review fixes
- rename opt_dont_trust_madvise to opt_trust_madvise
This fixes an incorrect debug-mode assert:
- T1 starts an arena stats update and reads stack_head from another thread's
cache bin, when that cache bin has 1 item in it.
- T2 allocates from that cache bin. The cache_bin's stack_head now points to a
NULL pointer, since the cache bin is empty.
- T1 Re-reads the cache_bin's stack_head to perform an assertion check (since it
previously saw that the bin was empty, whatever stack_head points to should be
non-NULL).
The additional overhead of the function-call setup and flags checking is
relatively small, but costs us the replication of the entire realloc pathway in
terms of size.
This hints to the compiler that it should care more about space than CPU (among
other things). In cases where the compiler lacks profile-guided information,
this can be a substantial space savings.
For now, we mark the mallctl or atexit driven profiling and stats functions that
take up the most space.
We do not fail on partial ctl path when the given `mib` array is
shorter than the given name, and we should keep the behavior the
same in the reverse case, which I feel is also the more natural way.
Now that we have flat bitmap bit counting functions, we can easily assert that
nfree is always correct. While we're tightening up this code, enforce
consistency on API boundaries as well.
This is no longer part of the "core" functionality; we only need the stub
implementations as an end-to-end test of hpdata + psset interactions when
metadata is being modified. Treat them accordingly.
Using an edata_t both for hugepages and the allocations within those hugepages
was convenient at first, but has outlived its usefulness. Representing
hugepages explicitly, with their own data structure, will make future
development easier.
With recent scalability improvements to the HPA, we're experimenting with much
lower arena counts; this gets annoying when trying to test across different
hardware configurations using only the narenas setting.
This was promised in the review of the introduction of geom_grow, but would have
been painful to do there because of the series that introduced it. Now that
those are comitted, renaming is easier.
At least one libc (musl) defines pthread_setname_np without defining
pthread_getname_np. Detect the presence of each individually, rather than
inferring both must be defined if set is.
In previous designs, this was intended to be a sort of cache that couldn't fail.
In the current design, we want to use it just as a contention reduction
mechanism. Rewrite it with those goals in mind.
This (experimental, undocumented) functionality can be used by users to track
various statistics of interest at a finer level of granularity than the thread.
Previously all the small size classes were cached. However this has downsides
-- particularly when page size is greater than 4K (e.g. iOS), which will result
in much higher SMALL_MAXCLASS.
This change allows tcache_max to be set to lower values, to better control
resources taken by tcache.
This functions more like the serial number strategy of the ecache and
hpa_central_t. Longer-lived slabs are more likely to continue to live for
longer in the future.
For locality reasons, tcache bins are integrated in TSD. Allowing all size
classes to be cached has little benefit, but takes up much thread local storage.
In addition, it complicates the layout which we try hard to optimize.
Without a lock held continuously between checking tcaches_past and incrementing
it, it's possible for two threads to go down manual creation path
simultaneously. If the number of tcaches is one less than the maximum, it's
possible for both to create a tcache and increment tcaches_past, with the second
thread returning a value larger than TCACHES_MAX.
This comes in handy when overriding earlier settings to test alternate ones. We
don't really include tests for this, but I claim that's OK here:
- It's fairly straightforward
- It's fairly hard to test well
- This entire code path is undocumented and mostly for our internal
experimentation in the first place.
- I tested manually.
This will be the centralized component of the coming hugepage allocator; the
source of larger chunks of memory from which smaller ones can be obtained.
This introduces a new sort of edata_t; a pageslab, and a set to manage them.
This is part of a series of a commits to implement a hugepage allocator; the
pageset will be per-arena, and track small page allocations requests within a
larger extent allocated from a centralized hugepage allocator.
This allows setting arenas per cpu dynamically, rather than forcing the user to
know the number of CPUs in advance if they want a particular CPU/space tradeoff.
The existing checks are good at finding such issues (on tcache flush), but not
so good at pinpointing them. Debug mode can find them, but sometimes debug mode
slows down a program so much that hard-to-hit bugs can take a long time to
crash.
This commit adds functionality to keep programs mostly on their fast paths,
while also checking every sized delete argument they get.
The sized dealloc checks called the generic safety_check_fail, and then called
abort. This means the failure case isn't mockable, hence not testable. Fix it
in anticipation of a coming diff.
This gives more accurate attribution of bytes and counts to stack traces,
without introducing backwards incompatibilities in heap-profile parsing tools.
We track the ideal reported (to the end user) number of bytes more carefully
inside core jemalloc. When dumping heap profiles, insteading of outputting our
counts directly, we output counts that will cause parsing tools to give a result
close to the value we want.
We retain the old version as an opt setting, to let users who are tracking
values on a per-component basis to keep their metrics stable until they decide
to switch.
For now, this is just a stub containing the ecaches, with no surrounding code
changed. Eventually all the core allocator bits will be moved in, in the
subsequent stack of commits.
In making these settings configurable, 634afc4124
unintentially changed a tuning parameter (reducing the "goal" max by a factor of
4). This commit undoes that change.
