Add ELF_DYNAMIC_AFTER_RELOC to allow target specific processing after
relocation.
For x86-64, add
#define DT_X86_64_PLT (DT_LOPROC + 0)
#define DT_X86_64_PLTSZ (DT_LOPROC + 1)
#define DT_X86_64_PLTENT (DT_LOPROC + 3)
1. DT_X86_64_PLT: The address of the procedure linkage table.
2. DT_X86_64_PLTSZ: The total size, in bytes, of the procedure linkage
table.
3. DT_X86_64_PLTENT: The size, in bytes, of a procedure linkage table
entry.
With the r_addend field of the R_X86_64_JUMP_SLOT relocation set to the
memory offset of the indirect branch instruction.
Define ELF_DYNAMIC_AFTER_RELOC for x86-64 to rewrite the PLT section
with direct branch after relocation when the lazy binding is disabled.
PLT rewrite is disabled by default since SELinux may disallow modifying
code pages and ld.so can't detect it in all cases. Use
$ export GLIBC_TUNABLES=glibc.cpu.plt_rewrite=1
to enable PLT rewrite with 32-bit direct jump at run-time or
$ export GLIBC_TUNABLES=glibc.cpu.plt_rewrite=2
to enable PLT rewrite with 32-bit direct jump and on APX processors with
64-bit absolute jump at run-time.
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Since malloc debug support moved to a different library
(libc_malloc_debug.so), the glibc.malloc.check requires preloading the
debug library to enable it. It means that suid-debug support has not
been working since 2.34.
To restore its support, it would require to add additional information
and parsing to where to find libc_malloc_debug.so.
It is one thing less that might change AT_SECURE binaries' behavior
due to environment configurations.
Checked on x86_64-linux-gnu.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
The latest implementations of memcpy are actually faster than the Falkor
implementations [1], so remove the falkor/phecda ifuncs for memcpy and
the now unused IS_FALKOR/IS_PHECDA defines.
[1] https://sourceware.org/pipermail/libc-alpha/2022-December/144227.html
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The PR_SET_VMA_ANON_NAME support is only enabled through a configurable
kernel switch, mainly because assigning a name to a
anonymous virtual memory area might prevent that area from being
merged with adjacent virtual memory areas.
For instance, with the following code:
void *p1 = mmap (NULL,
1024 * 4096,
PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS,
-1,
0);
void *p2 = mmap (p1 + (1024 * 4096),
1024 * 4096,
PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS,
-1,
0);
The kernel will potentially merge both mappings resulting in only one
segment of size 0x800000. If the segment is names with
PR_SET_VMA_ANON_NAME with different names, it results in two mappings.
Although this will unlikely be an issue for pthread stacks and malloc
arenas (since for pthread stacks the guard page will result in
a PROT_NONE segment, similar to the alignment requirement for the arena
block), it still might prevent the mmap memory allocated for detail
malloc.
There is also another potential scalability issue, where the prctl
requires
to take the mmap global lock which is still not fully fixed in Linux
[1] (for pthread stacks and arenas, it is mitigated by the stack
cached and the arena reuse).
So this patch disables anonymous mapping annotations as default and
add a new tunable, glibc.mem.decorate_maps, can be used to enable
it.
[1] https://lwn.net/Articles/906852/
Checked on x86_64-linux-gnu and aarch64-linux-gnu.
Reviewed-by: DJ Delorie <dj@redhat.com>
This patch enables the option to influence hwcaps used by PowerPC.
The environment variable, GLIBC_TUNABLES=glibc.cpu.hwcaps=-xxx,yyy,-zzz....,
can be used to enable CPU/ARCH feature yyy, disable CPU/ARCH feature xxx
and zzz, where the feature name is case-sensitive and has to match the ones
mentioned in the file{sysdeps/powerpc/dl-procinfo.c}.
Note that the hwcap tunables only used in the IFUNC selection.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Created tunable glibc.pthread.stack_hugetlb to control when hugepages
can be used for stack allocation.
In case THP are enabled and glibc.pthread.stack_hugetlb is set to
0, glibc will madvise the kernel not to use allow hugepages for stack
allocations.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Crossing 2GB boundaries with indirect calls and jumps can use more
branch prediction resources on Intel Golden Cove CPU (see the
"Misprediction for Branches >2GB" section in Intel 64 and IA-32
Architectures Optimization Reference Manual.) There is visible
performance improvement on workloads with many PLT calls when executable
and shared libraries are mmapped below 2GB. Add the Prefer_MAP_32BIT_EXEC
bit so that mmap will try to map executable or denywrite pages in shared
libraries with MAP_32BIT first.
NB: Prefer_MAP_32BIT_EXEC reduces bits available for address space
layout randomization (ASLR), which is always disabled for SUID programs
and can only be enabled by the tunable, glibc.cpu.prefer_map_32bit_exec,
or the environment variable, LD_PREFER_MAP_32BIT_EXEC. This works only
between shared libraries or between shared libraries and executables with
addresses below 2GB. PIEs are usually loaded at a random address above
4GB by the kernel.
When mcount overflows, no gmon.out file is generated, but no message is printed
to the user, leaving the user with no idea why, and thinking maybe there is
some bug - which is how BZ 27576 ended up being logged. Print a message to
stderr in this case so the user knows what is going on.
As a comment in sys/gmon.h acknowledges, the hardcoded MAXARCS value is too
small for some large applications, including the test case in that BZ. Rather
than increase it, add tunables to enable MINARCS and MAXARCS to be overridden
at runtime (glibc.gmon.minarcs and glibc.gmon.maxarcs). So if a user gets the
mcount overflow error, they can try increasing maxarcs (they might need to
increase minarcs too if the heuristic is wrong in their case.)
Note setting minarcs/maxarcs too large can cause monstartup to fail with an
out of memory error. If you set them large enough, it can cause an integer
overflow in calculating the buffer size. I haven't done anything to defend
against that - it would not generally be a security vulnerability, since these
tunables will be ignored in suid/sgid programs (due to the SXID_ERASE default),
and if you can set GLIBC_TUNABLES in the environment of a process, you can take
it over anyway (LD_PRELOAD, LD_LIBRARY_PATH, etc). I thought about modifying
the code of monstartup to defend against integer overflows, but doing so is
complicated, and I realise the existing code is susceptible to them even prior
to this change (e.g. try passing a pathologically large highpc argument to
monstartup), so I decided just to leave that possibility in-place.
Add a test case which demonstrates mcount overflow and the tunables.
Document the new tunables in the manual.
Signed-off-by: Simon Kissane <skissane@gmail.com>
Reviewed-by: DJ Delorie <dj@redhat.com>
This patch enables the option to influence hwcaps and stfle bits used
by the s390 specific ifunc-resolvers. The currently x86-specific
tunable glibc.cpu.hwcaps is also used on s390x to achieve the task. In
addition the user can also set a CPU arch-level like z13 instead of
single HWCAP and STFLE features.
Note that the tunable only handles the features which are really used
in the IFUNC-resolvers. All others are ignored as the values are only
used inside glibc. Thus we can influence:
- HWCAP_S390_VXRS (z13)
- HWCAP_S390_VXRS_EXT (z14)
- HWCAP_S390_VXRS_EXT2 (z15)
- STFLE_MIE3 (z15)
The influenced hwcap/stfle-bits are stored in the s390-specific
cpu_features struct which also contains reserved fields for future
usage.
The ifunc-resolvers and users of stfle bits are adjusted to use the
information from cpu_features struct.
On 31bit, the ELF_MACHINE_IRELATIVE macro is now also defined.
Otherwise the new ifunc-resolvers segfaults as they depend on
the not yet processed_rtld_global_ro@GLIBC_PRIVATE relocation.
The new asymmetric mode is available when HWCAP2_MTE3 is set (support is
available), bit2 is set in the tunable (user request per application),
and the system is configured such that the asymmetric mode is preferred over
sync or async (per-cpu system-wide setting).
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
The lower-bound (16448) and upper-bound (SIZE_MAX / 16) are assumed
by memmove-vec-unaligned-erms.
The lower-bound is needed because memmove-vec-unaligned-erms unrolls
the loop aggressively in the L(large_memset_4x) case.
The upper-bound is needed because memmove-vec-unaligned-erms
right-shifts the value of `x86_non_temporal_threshold` by
LOG_4X_MEMCPY_THRESH (4) which without a bound may overflow.
The lack of lower-bound can be a correctness issue. The lack of
upper-bound cannot.
With the morecore hook removed, there is not easy way to provide huge
pages support on with glibc allocator without resorting to transparent
huge pages. And some users and programs do prefer to use the huge pages
directly instead of THP for multiple reasons: no splitting, re-merging
by the VM, no TLB shootdowns for running processes, fast allocation
from the reserve pool, no competition with the rest of the processes
unlike THP, no swapping all, etc.
This patch extends the 'glibc.malloc.hugetlb' tunable: the value
'2' means to use huge pages directly with the system default size,
while a positive value means and specific page size that is matched
against the supported ones by the system.
Currently only memory allocated on sysmalloc() is handled, the arenas
still uses the default system page size.
To test is a new rule is added tests-malloc-hugetlb2, which run the
addes tests with the required GLIBC_TUNABLE setting. On systems without
a reserved huge pages pool, is just stress the mmap(MAP_HUGETLB)
allocation failure. To improve test coverage it is required to create
a pool with some allocated pages.
Checked on x86_64-linux-gnu.
Reviewed-by: DJ Delorie <dj@redhat.com>
Linux Transparent Huge Pages (THP) current supports three different
states: 'never', 'madvise', and 'always'. The 'never' is
self-explanatory and 'always' will enable THP for all anonymous
pages. However, 'madvise' is still the default for some system and
for such case THP will be only used if the memory range is explicity
advertise by the program through a madvise(MADV_HUGEPAGE) call.
To enable it a new tunable is provided, 'glibc.malloc.hugetlb',
where setting to a value diffent than 0 enables the madvise call.
This patch issues the madvise(MADV_HUGEPAGE) call after a successful
mmap() call at sysmalloc() with sizes larger than the default huge
page size. The madvise() call is disable is system does not support
THP or if it has the mode set to "never" and on Linux only support
one page size for THP, even if the architecture supports multiple
sizes.
To test is a new rule is added tests-malloc-hugetlb1, which run the
addes tests with the required GLIBC_TUNABLE setting.
Checked on x86_64-linux-gnu.
Reviewed-by: DJ Delorie <dj@redhat.com>
The default has to change eventually, and there are no known failures
that require a delay.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
This tunable allows applications to register the rseq area instead
of glibc.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
This second patch contains the actual implementation of a new sorting algorithm
for shared objects in the dynamic loader, which solves the slow behavior that
the current "old" algorithm falls into when the DSO set contains circular
dependencies.
The new algorithm implemented here is simply depth-first search (DFS) to obtain
the Reverse-Post Order (RPO) sequence, a topological sort. A new l_visited:1
bitfield is added to struct link_map to more elegantly facilitate such a search.
The DFS algorithm is applied to the input maps[nmap-1] backwards towards
maps[0]. This has the effect of a more "shallow" recursion depth in general
since the input is in BFS. Also, when combined with the natural order of
processing l_initfini[] at each node, this creates a resulting output sorting
closer to the intuitive "left-to-right" order in most cases.
Another notable implementation adjustment related to this _dl_sort_maps change
is the removing of two char arrays 'used' and 'done' in _dl_close_worker to
represent two per-map attributes. This has been changed to simply use two new
bit-fields l_map_used:1, l_map_done:1 added to struct link_map. This also allows
discarding the clunky 'used' array sorting that _dl_sort_maps had to sometimes
do along the way.
Tunable support for switching between different sorting algorithms at runtime is
also added. A new tunable 'glibc.rtld.dynamic_sort' with current valid values 1
(old algorithm) and 2 (new DFS algorithm) has been added. At time of commit
of this patch, the default setting is 1 (old algorithm).
Signed-off-by: Chung-Lin Tang <cltang@codesourcery.com>
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
All references to libraries in the manual are without the .so prefix,
so do the same for libc_malloc_debug.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Remove all malloc hook uses from core malloc functions and move it
into a new library libc_malloc_debug.so. With this, the hooks now no
longer have any effect on the core library.
libc_malloc_debug.so is a malloc interposer that needs to be preloaded
to get hooks functionality back so that the debugging features that
depend on the hooks, i.e. malloc-check, mcheck and mtrace work again.
Without the preloaded DSO these debugging features will be nops.
These features will be ported away from hooks in subsequent patches.
Similarly, legacy applications that need hooks functionality need to
preload libc_malloc_debug.so.
The symbols exported by libc_malloc_debug.so are maintained at exactly
the same version as libc.so.
Finally, static binaries will no longer be able to use malloc
debugging features since they cannot preload the debugging DSO.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Tested-by: Carlos O'Donell <carlos@redhat.com>
The tunable will not work with *any* non-zero tunable value since its
list of allowed values is 0-3. Fix the documentation to reflect that.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The valgrind/helgrind test suite needs a way to make stack dealloction
more prompt, and this feature seems to be generally useful.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
This patch optimizes the performance of memcpy/memmove for A64FX [1]
which implements ARMv8-A SVE and has L1 64KB cache per core and L2 8MB
cache per NUMA node.
The performance optimization makes use of Scalable Vector Register
with several techniques such as loop unrolling, memory access
alignment, cache zero fill, and software pipelining.
SVE assembler code for memcpy/memmove is implemented as Vector Length
Agnostic code so theoretically it can be run on any SOC which supports
ARMv8-A SVE standard.
We confirmed that all testcases have been passed by running 'make
check' and 'make xcheck' not only on A64FX but also on ThunderX2.
And also we confirmed that the SVE 512 bit vector register performance
is roughly 4 times better than Advanced SIMD 128 bit register and 8
times better than scalar 64 bit register by running 'make bench'.
[1] https://github.com/fujitsu/A64FX
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
Reviewed-by: Szabolcs Nagy <Szabolcs.Nagy@arm.com>
Cover key corner cases (e.g., whether errno is set) that are well
settled in glibc, fix some examples to avoid integer overflow, and
update some other dated examples (code needed for K&R C, e.g.).
* manual/charset.texi (Non-reentrant String Conversion):
* manual/filesys.texi (Symbolic Links):
* manual/memory.texi (Allocating Cleared Space):
* manual/socket.texi (Host Names):
* manual/string.texi (Concatenating Strings):
* manual/users.texi (Setting Groups):
Use reallocarray instead of realloc, to avoid integer overflow issues.
* manual/filesys.texi (Scanning Directory Content):
* manual/memory.texi (The GNU Allocator, Hooks for Malloc):
* manual/tunables.texi:
Use code font for 'malloc' instead of roman font.
(Symbolic Links): Don't assume readlink return value fits in 'int'.
* manual/memory.texi (Memory Allocation and C, Basic Allocation)
(Malloc Examples, Alloca Example):
* manual/stdio.texi (Formatted Output Functions):
* manual/string.texi (Concatenating Strings, Collation Functions):
Omit pointer casts that are needed only in ancient K&R C.
* manual/memory.texi (Basic Allocation):
Say that malloc sets errno on failure.
Say "convert" rather than "cast", since casts are no longer needed.
* manual/memory.texi (Basic Allocation):
* manual/string.texi (Concatenating Strings):
In examples, use C99 declarations after statements for brevity.
* manual/memory.texi (Malloc Examples): Add portability notes for
malloc (0), errno setting, and PTRDIFF_MAX.
(Changing Block Size): Say that realloc (p, 0) acts like
(p ? (free (p), NULL) : malloc (0)).
Add xreallocarray example, since other examples can use it.
Add portability notes for realloc (0, 0), realloc (p, 0),
PTRDIFF_MAX, and improve notes for reallocating to the same size.
(Allocating Cleared Space): Reword now-confusing discussion
about replacement, and xref "Replacing malloc".
* manual/stdio.texi (Formatted Output Functions):
Don't assume message size fits in 'int'.
* manual/string.texi (Concatenating Strings):
Fix undefined behavior involving arithmetic on a freed pointer.
Add a new glibc tunable: mem.tagging. This is a decimal constant in
the range 0-255 but used as a bit-field.
Bit 0 enables use of tagged memory in the malloc family of functions.
Bit 1 enables precise faulting of tag failure on platforms where this
can be controlled.
Other bits are currently unused, but if set will cause memory tag
checking for the current process to be enabled in the kernel.
The __x86_shared_non_temporal_threshold determines when memcpy on x86
uses non_temporal stores to avoid pushing other data out of the last
level cache.
This patch proposes to revert the calculation change made by H.J. Lu's
patch of June 2, 2017.
H.J. Lu's patch selected a threshold suitable for a single thread
getting maximum performance. It was tuned using the single threaded
large memcpy micro benchmark on an 8 core processor. The last change
changes the threshold from using 3/4 of one thread's share of the
cache to using 3/4 of the entire cache of a multi-threaded system
before switching to non-temporal stores. Multi-threaded systems with
more than a few threads are server-class and typically have many
active threads. If one thread consumes 3/4 of the available cache for
all threads, it will cause other active threads to have data removed
from the cache. Two examples show the range of the effect. John
McCalpin's widely parallel Stream benchmark, which runs in parallel
and fetches data sequentially, saw a 20% slowdown with this patch on
an internal system test of 128 threads. This regression was discovered
when comparing OL8 performance to OL7. An example that compares
normal stores to non-temporal stores may be found at
https://vgatherps.github.io/2018-09-02-nontemporal/. A simple test
shows performance loss of 400 to 500% due to a failure to use
nontemporal stores. These performance losses are most likely to occur
when the system load is heaviest and good performance is critical.
The tunable x86_non_temporal_threshold can be used to override the
default for the knowledgable user who really wants maximum cache
allocation to a single thread in a multi-threaded system.
The manual entry for the tunable has been expanded to provide
more information about its purpose.
modified: sysdeps/x86/cacheinfo.c
modified: manual/tunables.texi
On some targets static TLS surplus area can be used opportunistically
for dynamically loaded modules such that the TLS access then becomes
faster (TLSDESC and powerpc TLS optimization). However we don't want
all surplus TLS to be used for this optimization because dynamically
loaded modules with initial-exec model TLS can only use surplus TLS.
The new contract for surplus static TLS use is:
- libc.so can have up to 192 bytes of IE TLS,
- other system libraries together can have up to 144 bytes of IE TLS.
- Some "optional" static TLS is available for opportunistic use.
The optional TLS is now tunable: rtld.optional_static_tls, so users
can directly affect the allocated static TLS size. (Note that module
unloading with dlclose does not reclaim static TLS. After the optional
TLS runs out, TLS access is no longer optimized to use static TLS.)
The default setting of rtld.optional_static_tls is 512 so the surplus
TLS is 3*192 + 4*144 + 512 = 1664 by default, the same as before.
Fixes BZ #25051.
Tested on aarch64-linux-gnu and x86_64-linux-gnu.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The new static TLS surplus size computation is
surplus_tls = 192 * (nns-1) + 144 * nns + 512
where nns is controlled via the rtld.nns tunable. This commit
accounts audit modules too so nns = rtld.nns + audit modules.
rtld.nns should only include the namespaces required by the
application, namespaces for audit modules are accounted on top
of that so audit modules don't use up the static TLS that is
reserved for the application. This allows loading many audit
modules without tuning rtld.nns or using up static TLS, and it
fixes
FAIL: elf/tst-auditmany
Note that DL_NNS is currently a hard upper limit for nns, and
if rtld.nns + audit modules go over the limit that's a fatal
error. By default rtld.nns is 4 which allows 12 audit modules.
Counting the audit modules is based on existing audit string
parsing code, we cannot use GLRO(dl_naudit) before the modules
are actually loaded.
TLS_STATIC_SURPLUS is 1664 bytes currently which is not enough to
support DL_NNS (== 16) number of dynamic link namespaces, if we
assume 192 bytes of TLS are reserved for libc use and 144 bytes
are reserved for other system libraries that use IE TLS.
A new tunable is introduced to control the number of supported
namespaces and to adjust the surplus static TLS size as follows:
surplus_tls = 192 * (rtld.nns-1) + 144 * rtld.nns + 512
The default is rtld.nns == 4 and then the surplus TLS size is the
same as before, so the behaviour is unchanged by default. If an
application creates more namespaces than the rtld.nns setting
allows, then it is not guaranteed to work, but the limit is not
checked. So existing usage will continue to work, but in the
future if an application creates more than 4 dynamic link
namespaces then the tunable will need to be set.
In this patch DL_NNS is a fixed value and provides a maximum to
the rtld.nns setting.
Static linking used fixed 2048 bytes surplus TLS, this is changed
so the same contract is used as for dynamic linking. With static
linking DL_NNS == 1 so rtld.nns tunable is forced to 1, so by
default the surplus TLS is reduced to 144 + 512 = 656 bytes. This
change is not expected to cause problems.
Tested on aarch64-linux-gnu and x86_64-linux-gnu.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Add x86_rep_movsb_threshold and x86_rep_stosb_threshold to tunables
to update thresholds for "rep movsb" and "rep stosb" at run-time.
Note that the user specified threshold for "rep movsb" smaller than
the minimum threshold will be ignored.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Kunpeng processer is a 64-bit Arm-compatible CPU released by Huawei,
and we have already signed a copyright assignement with the FSF.
This patch adds its to cpu list, and related macro for IFUNC.
Checked on aarch64-linux-gnu.
Reviewed-by: Szabolcs Nagy <Szabolcs.Nagy@arm.com>
Change the tcache->counts[] entries to uint16_t - this removes
the limit set by char and allows a larger tcache. Remove a few
redundant asserts.
bench-malloc-thread with 4 threads is ~15% faster on Cortex-A72.
Reviewed-by: DJ Delorie <dj@redhat.com>
* malloc/malloc.c (MAX_TCACHE_COUNT): Increase to UINT16_MAX.
(tcache_put): Remove redundant assert.
(tcache_get): Remove redundant asserts.
(__libc_malloc): Check tcache count is not zero.
* manual/tunables.texi (glibc.malloc.tcache_count): Update maximum.
The tcache counts[] array is a char, which has a very small range and thus
may overflow. When setting tcache_count tunable, there is no overflow check.
However the tunable must not be larger than the maximum value of the tcache
counts[] array, otherwise it can overflow when filling the tcache.
[BZ #24531]
* malloc/malloc.c (MAX_TCACHE_COUNT): New define.
(do_set_tcache_count): Only update if count is small enough.
* manual/tunables.texi (glibc.malloc.tcache_count): Document max value.
Emag is a 64-bit CPU core released by AmpereComputing.
Add its name to cpu list, and corresponding macro as utilities for
later IFUNC dispatch.
* manual/tunables.texi (Tunable glibc.cpu.name): Add emag.
* sysdeps/unix/sysv/linux/aarch64/cpu-features.c (cpu_list):
Add emag.
* sysdeps/unix/sysv/linux/aarch64/cpu-features.h (IS_EMAG):
New macro.
Add Ares to the midr_el0 list and support ifunc dispatch. Since Ares
supports 2 128-bit loads/stores, use Neon registers for memcpy by
selecting __memcpy_falkor by default (we should rename this to
__memcpy_simd or similar).
* manual/tunables.texi (glibc.cpu.name): Add ares tunable.
* sysdeps/aarch64/multiarch/memcpy.c (__libc_memcpy): Use
__memcpy_falkor for ares.
* sysdeps/unix/sysv/linux/aarch64/cpu-features.h (IS_ARES):
Add new define.
* sysdeps/unix/sysv/linux/aarch64/cpu-features.c (cpu_list):
Add ares cpu.
This patch does not have any functionality change, we only provide a spin
count tunes for pthread adaptive spin mutex. The tunable
glibc.pthread.mutex_spin_count tunes can be used by system administrator to
squeeze system performance according to different hardware capabilities and
workload characteristics.
The maximum value of spin count is limited to 32767 to avoid the overflow
of mutex->__data.__spins variable with the possible type of short in
pthread_mutex_lock ().
The default value of spin count is set to 100 with the reference to the
previous number of times of spinning via trylock. This value would be
architecture-specific and can be tuned with kinds of benchmarks to fit most
cases in future.
I would extend my appreciation sincerely to H.J.Lu for his help to refine
this patch series.
* manual/tunables.texi (POSIX Thread Tunables): New node.
* nptl/Makefile (libpthread-routines): Add pthread_mutex_conf.
* nptl/nptl-init.c: Include pthread_mutex_conf.h
(__pthread_initialize_minimal_internal) [HAVE_TUNABLES]: Call
__pthread_tunables_init.
* nptl/pthreadP.h (MAX_ADAPTIVE_COUNT): Remove.
(max_adaptive_count): Define.
* nptl/pthread_mutex_conf.c: New file.
* nptl/pthread_mutex_conf.h: New file.
* sysdeps/generic/adaptive_spin_count.h: New file.
* sysdeps/nptl/dl-tunables.list: New file.
* nptl/pthread_mutex_lock.c (__pthread_mutex_lock): Use
max_adaptive_count () not MAX_ADAPTIVE_COUNT.
* nptl/pthread_mutex_timedlock.c (__pthrad_mutex_timedlock):
Likewise.
Suggested-by: Andi Kleen <andi.kleen@intel.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Signed-off-by: Kemi.wang <kemi.wang@intel.com>
The glibc.tune namespace is vaguely named since it is a 'tunable', so
give it a more specific name that describes what it refers to. Rename
the tunable namespace to 'cpu' to more accurately reflect what it
encompasses. Also rename glibc.tune.cpu to glibc.cpu.name since
glibc.cpu.cpu is weird.
* NEWS: Mention the change.
* elf/dl-tunables.list: Rename tune namespace to cpu.
* sysdeps/powerpc/dl-tunables.list: Likewise.
* sysdeps/x86/dl-tunables.list: Likewise.
* sysdeps/aarch64/dl-tunables.list: Rename tune.cpu to
cpu.name.
* elf/dl-hwcaps.c (_dl_important_hwcaps): Adjust.
* elf/dl-hwcaps.h (GET_HWCAP_MASK): Likewise.
* manual/README.tunables: Likewise.
* manual/tunables.texi: Likewise.
* sysdeps/powerpc/cpu-features.c: Likewise.
* sysdeps/unix/sysv/linux/aarch64/cpu-features.c
(init_cpu_features): Likewise.
* sysdeps/x86/cpu-features.c: Likewise.
* sysdeps/x86/cpu-features.h: Likewise.
* sysdeps/x86/cpu-tunables.c: Likewise.
* sysdeps/x86_64/Makefile: Likewise.
* sysdeps/x86/dl-cet.c: Likewise.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
On POWER8, unaligned memory accesses to cached memory has little impact
on performance as opposed to its ancestors.
It is disabled by default and will only be available when the tunable
glibc.tune.cached_memopt is set to 1.
__memcpy_power8_cached __memcpy_power7
============================================================
max-size=4096: 33325.70 ( 12.65%) 38153.00
max-size=8192: 32878.20 ( 11.17%) 37012.30
max-size=16384: 33782.20 ( 11.61%) 38219.20
max-size=32768: 33296.20 ( 11.30%) 37538.30
max-size=65536: 33765.60 ( 10.53%) 37738.40
* manual/tunables.texi (Hardware Capability Tunables): Document
glibc.tune.cached_memopt.
* sysdeps/powerpc/cpu-features.c: New file.
* sysdeps/powerpc/cpu-features.h: New file.
* sysdeps/powerpc/dl-procinfo.c [!IS_IN(ldconfig)]: Add
_dl_powerpc_cpu_features.
* sysdeps/powerpc/dl-tunables.list: New file.
* sysdeps/powerpc/ldsodefs.h: Include cpu-features.h.
* sysdeps/powerpc/powerpc32/power4/multiarch/init-arch.h
(INIT_ARCH): Initialize use_aligned_memopt.
* sysdeps/powerpc/powerpc64/dl-machine.h [defined(SHARED &&
IS_IN(rtld))]: Restrict dl_platform_init availability and
initialize CPU features used by tunables.
* sysdeps/powerpc/powerpc64/multiarch/Makefile (sysdep_routines):
Add memcpy-power8-cached.
* sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c: Add
__memcpy_power8_cached.
* sysdeps/powerpc/powerpc64/multiarch/memcpy.c: Likewise.
* sysdeps/powerpc/powerpc64/multiarch/memcpy-power8-cached.S:
New file.
Reviewed-by: Rajalakshmi Srinivasaraghavan <raji@linux.vnet.ibm.com>
This patch adds several new tunables to control the behavior of
elision on supported platforms[1]. Since elision now depends
on tunables, we should always *compile* with elision enabled,
and leave the code disabled, but available for runtime
selection. This gives us *much* better compile-time testing of
the existing code to avoid bit-rot[2].
Tested on ppc, ppc64, ppc64le, s390x and x86_64.
[1] This part of the patch was initially proposed by
Paul Murphy but was "staled" because the framework have changed
since the patch was originally proposed:
https://patchwork.sourceware.org/patch/10342/
[2] This part of the patch was inititally proposed as a RFC by
Carlos O'Donnell. Make sense to me integrate this on the patch:
https://sourceware.org/ml/libc-alpha/2017-05/msg00335.html
* elf/dl-tunables.list: Add elision parameters.
* manual/tunables.texi: Add entries about elision tunable.
* sysdeps/unix/sysv/linux/powerpc/elision-conf.c:
Add callback functions to dynamically enable/disable elision.
Add multiple callbacks functions to set elision parameters.
Deleted __libc_enable_secure check.
* sysdeps/unix/sysv/linux/s390/elision-conf.c: Likewise.
* sysdeps/unix/sysv/linux/x86/elision-conf.c: Likewise.
* configure: Regenerated.
* configure.ac: Option enable_lock_elision was deleted.
* config.h.in: ENABLE_LOCK_ELISION flag was deleted.
* config.make.in: Remove references to enable_lock_elision.
* manual/install.texi: Elision configure option was removed.
* INSTALL: Regenerated to remove enable_lock_elision.
* nptl/Makefile:
Disable elision so it can verify error case for destroying a mutex.
* sysdeps/powerpc/nptl/elide.h:
Cleanup ENABLE_LOCK_ELISION check.
Deleted macros for the case when ENABLE_LOCK_ELISION was not defined.
* sysdeps/s390/configure: Regenerated.
* sysdeps/s390/configure.ac: Remove references to enable_lock_elision..
* nptl/tst-mutex8.c:
Deleted all #ifndef ENABLE_LOCK_ELISION from the test.
* sysdeps/powerpc/powerpc32/sysdep.h:
Deleted all ENABLE_LOCK_ELISION checks.
* sysdeps/powerpc/powerpc64/sysdep.h: Likewise.
* sysdeps/powerpc/sysdep.h: Likewise.
* sysdeps/s390/nptl/bits/pthreadtypes-arch.h: Likewise.
* sysdeps/unix/sysv/linux/powerpc/force-elision.h: Likewise.
* sysdeps/unix/sysv/linux/s390/elision-conf.h: Likewise.
* sysdeps/unix/sysv/linux/s390/force-elision.h: Likewise.
* sysdeps/unix/sysv/linux/s390/lowlevellock.h: Likewise.
* sysdeps/unix/sysv/linux/s390/Makefile: Remove references to
enable-lock-elision.
Reviewed-by: Tulio Magno Quites Machado Filho <tuliom@linux.vnet.ibm.com>
* manual/tunables.texi (glibc.tune.cpu): Add thunderx2t99 and
thunderx2t99p1 to list of cpu names.
* sysdeps/unix/sysv/linux/aarch64/cpu-features.c (cpu_list):
Add thunderx2t99 and thunderx2t99p1 entries to cpu_list.
