This patch enables libmvec on AArch64. The proposed change is mainly
implementing build infrastructure to add the new routines to ABI,
tests and benchmarks. I have demonstrated how this all fits together
by adding implementations for vector cos, in both single and double
precision, targeting both Advanced SIMD and SVE.
The implementations of the routines themselves are just loops over the
scalar routine from libm for now, as we are more concerned with
getting the plumbing right at this point. We plan to contribute vector
routines from the Arm Optimized Routines repo that are compliant with
requirements described in the libmvec wiki.
Building libmvec requires minimum GCC 10 for SVE ACLE. To avoid raising
the minimum GCC by such a big jump, we allow users to disable libmvec
if their compiler is too old.
Note that at this point users have to manually call the vector math
functions. This seems to be acceptable to some downstream users.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
dev_t are 64bit on Linux ports, so better increase their size on 64bit
Hurd. It happens that this helps with BZ 23084 there: st_dev has type fsid_t
(quad) and is specified by POSIX to have type dev_t. Making dev_t 64bit
makes these match.
The standards want msg_lspid/msg_lrpid/shm_cpid/shm_lpid to be pid_t, see BZ
23083 and 23085.
We can leave them __rpc_pid_t on i386 for ABI compatibility, but avoid
hitting the issue on 64bit.
The standards want uid/cuid to be uid_t, gid/cgid to be gid_t and mode to be
mode_t, see BZ 23082.
We can leave them short ints on i386 for ABI compatibility, but avoid
hitting the issue on 64bit.
bits/ipc.h ends up being exactly the same in sysdeps/gnu/ and
sysdeps/unix/sysv/linux/, so remove the latter.
The standards want l_type and l_whence to be short ints, see BZ 23081.
We can leave them ints on i386 for ABI compatibility, but avoid hitting the
issue on 64bit.
These were created by creating stub files, running 'make update-abi',
and reviewing the results.
Also, set baseline ABI to GLIBC_2.38, the (upcoming) first glibc
release to first have x86_64-gnu support.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
If we're trying to interrupt an interruptible RPC, but the server fails
to respond to our __interrupt_operation () call, we instead destroy the
reply port we were expecting the reply to the RPC on.
Instead of deallocating the name completely, replace it with a dead
name, so the name won't get reused for some other right, and deallocate
it in _hurd_intr_rpc_mach_msg once we return from the signal handler.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230429201822.2605207-4-bugaevc@gmail.com>
We make lib{mach,hurd}user.so only call __mig_strlen which can be
relocated before libc.so is relocated, similar to what is done with
__mig_memcpy.
Message-Id: <ZE8DTRDpY2hpPZlJ@jupiter.tail36e24.ts.net>
Normally, in static builds, the first code that runs is _start, in e.g.
sysdeps/x86_64/start.S, which quickly calls __libc_start_main, passing
it the argv etc. Among the first things __libc_start_main does is
initializing the tunables (based on env), then CPU features, and then
calls _dl_relocate_static_pie (). Specifically, this runs ifunc
resolvers to pick, based on the CPU features discovered earlier, the
most suitable implementation of "string" functions such as memcpy.
Before that point, calling memcpy (or other ifunc-resolved functions)
will not work.
In the Hurd port, things are more complex. In order to get argv/env for
our process, glibc normally needs to do an RPC to the exec server,
unless our args/env are already located on the stack (which is what
happens to bootstrap processes spawned by GNU Mach). Fetching our
argv/env from the exec server has to be done before the call to
__libc_start_main, since we need to know what our argv/env are to pass
them to __libc_start_main.
On the other hand, the implementation of the RPC (and other initial
setup needed on the Hurd before __libc_start_main can be run) is not
very trivial. In particular, it may (and on x86_64, will) use memcpy.
But as described above, calling memcpy before __libc_start_main can not
work, since the GOT entry for it is not yet initialized at that point.
Work around this by pre-filling the GOT entry with the baseline version
of memcpy, __memcpy_sse2_unaligned. This makes it possible for early
calls to memcpy to just work. The initial value of the GOT entry is
unused on x86_64, and changing it won't interfere with the relocation
being performed later: once _dl_relocate_static_pie () is called, the
baseline version will get replaced with the most suitable one, and that
is what subsequent calls of memcpy are going to call.
Checked on x86_64-gnu.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230429201822.2605207-6-bugaevc@gmail.com>
Checked on x86_64-gnu.
[samuel.thibault@ens-lyon.org: Restored same comments as on i386]
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230429201822.2605207-3-bugaevc@gmail.com>
If any of the early boot-up tasks calls exit () or returns from main (),
terminate it properly instead of crashing on trying to dereference
_hurd_ports and getting forcibly terminated by the kernel.
We sadly cannot make the __USEPORT macro do the check for _hurd_ports
being unset, because it evaluates to the value of the expression
provided as the second argument, and that can be of any type; so there
is no single suitable fallback value for the macro to evaluate to in
case _hurd_ports is unset. Instead, each use site that wants to care for
this case will have to do its own checking.
Checked on x86_64-gnu.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230429131354.2507443-4-bugaevc@gmail.com>
There are reports for hang in __check_pf:
https://github.com/JoeDog/siege/issues/4
It is reproducible only under specific configurations:
1. Large number of cores (>= 64) and large number of threads (> 3X of
the number of cores) with long lived socket connection.
2. Low power (frequency) mode.
3. Power management is enabled.
While holding lock, __check_pf calls make_request which calls __sendto
and __recvmsg. Since __sendto and __recvmsg are cancellation points,
lock held by __check_pf won't be released and can cause deadlock when
thread cancellation happens in __sendto or __recvmsg. Add a cancellation
cleanup handler for __check_pf to unlock the lock when cancelled by
another thread. This fixes BZ #20975 and the siege hang issue.
In some cases, we do not want to go through the resolver for function
calls. For example, functions with vector arguments will use vector
registers to pass arguments. In the resolver, we do not save/restore the
vector argument registers for lazy binding efficiency. To avoid ruining
the vector arguments, functions with vector arguments will not go
through the resolver.
To achieve the goal, we will annotate the function symbols with
STO_RISCV_VARIANT_CC flag and add DT_RISCV_VARIANT_CC tag in the dynamic
section. In the first pass on PLT relocations, we do not set up to call
_dl_runtime_resolve. Instead, we resolve the functions directly.
Signed-off-by: Hsiangkai Wang <kai.wang@sifive.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com>
Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
Link: https://inbox.sourceware.org/libc-alpha/20230314162512.35802-1-kito.cheng@sifive.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
The build of glibc for i686-gnu has been failing for a while with GCC
mainline / GCC 13:
../sysdeps/mach/hurd/getcwd.c: In function '__hurd_canonicalize_directory_name_internal':
../sysdeps/mach/hurd/getcwd.c:242:48: error: pointer 'file_name' may be used after 'realloc' [-Werror=use-after-free]
242 | file_namep = &buf[file_namep - file_name + size / 2];
| ~~~~~~~~~~~^~~~~~~~~~~
../sysdeps/mach/hurd/getcwd.c:236:25: note: call to 'realloc' here
236 | buf = realloc (file_name, size);
| ^~~~~~~~~~~~~~~~~~~~~~~~~
Fix by doing the subtraction before the reallocation.
Tested with build-many-glibcs.py for i686-gnu.
[samuel.thibault@ens-lyon.rg: Removed mention of this being a bug]
Message-Id: <18587337-7815-4056-ebd0-724df262d591@codesourcery.com>
As fixed in 0822e3552a ("hurd: Don't pass FD_CLOEXEC in CMSG_DATA"),
senders currently don't have any flag to pass. We shouldn't blindly take
random flags that senders could be erroneously giving us.
This is a new flag that can be passed to recvmsg () to make it
atomically set the CLOEXEC flag on all the file descriptors received
using the SCM_RIGHTS mechanism. This is useful for all the same reasons
that the other XXX_CLOEXEC flags are useful: namely, it provides
atomicity with respect to another thread of the same process calling
(fork and then) exec at the same time.
This flag is already supported on Linux and FreeBSD. The flag's value,
0x40000, is choosen to match FreeBSD's.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230423160548.126576-2-bugaevc@gmail.com>
The flags are used by _hurd_intern_fd, which takes O_* flags, not FD_*.
Also, it is of no concern to the receiving process whether or not
the sender process wants to close its copy of sent file descriptor
upon exec, and it should not influence whether or not the received
file descriptor gets the FD_CLOEXEC flag set in the receiving process.
The latter should in fact be dependent on the MSG_CMSG_CLOEXEC flag
being passed to the recvmsg () call, which is going to be implemented
in the following commit.
Fixes 344e755248
"hurd: Support sending file descriptors over Unix sockets"
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
This makes the prefer_map_32bit_exec tunable no longer Linux-specific.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230423215526.346009-4-bugaevc@gmail.com>
This is a flag that can be passed to mmap () to request that the mapping
being established should be located in the lower 2 GB area of the
address space, so only the lower 31 (not 32) bits can be set in its
address, and the address can be represented as a 32-bit integer without
truncating it.
This flag is intended to be compatible with Linux, FreeBSD, and Darwin
flags of the same name. Out of those systems, it appears Linux and
FreeBSD take MAP_32BIT to mean "map 31 bit", whereas Darwin allows the
32nd bit to be set in the address as well. The Hurd follows Linux and
FreeBSD behavior.
Unlike on those systems, on the Hurd MAP_32BIT is defined on all
supported architectures (which currently are only i386 and x86_64).
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230423215526.346009-1-bugaevc@gmail.com>
When opening a temporary file without O_CLOEXEC we risk leaking the
file descriptor if another thread calls (fork and then) exec while we
have the fd open. Fix this by consistently passing O_CLOEXEC everywhere
where we open a file for internal use (and not to return it to the user,
in which case the API defines whether or not the close-on-exec flag
shall be set on the returned fd).
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230419160207.65988-4-bugaevc@gmail.com>
Properly differentiate between setting up the real TLS with
TLS_INIT_TP, and setting up the early TLS (__init1_tcbhead) in static
builds. In the latter case, don't yet migrate the reply port into the
TCB, and don't yet set __libc_tls_initialized to 1.
This also lets us move the __init1_desc assignment inside
_hurd_tls_init ().
Fixes cd019ddd89
"hurd: Don't leak __hurd_reply_port0"
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
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>
We must not use the user's reply port (scp->sc_reply_port) for any of
our own RPCs, otherwise various things break. So, use MACH_PORT_DEAD as
a reply port when destroying our reply port, and make sure to do this
after _hurd_sigstate_unlock (), which may do a gsync_wake () RPC.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Optimize the fast paths (x < y) and (x/y < 2^12). Delay handling of special
cases to reduce the number of instructions executed before the fast paths.
Performance improvements for fmod:
Skylake Zen2 Neoverse V1
subnormals 11.8% 4.2% 11.5%
normal 3.9% 0.01% -0.5%
close-exponents 6.3% 5.6% 19.4%
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
When glibc is built as a shared library, TLS is always initialized by
the call of TLS_INIT_TP () macro made inside the dynamic loader, prior
to running the main program (see dl-call_tls_init_tp.h). We can take
advantage of this: we know for sure that __LIBC_NO_TLS () will evaluate
to 0 in all other cases, so let the compiler know that explicitly too.
Also, only define _hurd_tls_init () and TLS_INIT_TP () under the same
conditions (either !SHARED or inside rtld), to statically assert that
this is the case.
Other than a microoptimization, this also helps with avoiding awkward
sharing of the __libc_tls_initialized variable between ld.so and libc.so
that we would have to do otherwise -- we know for sure that no sharing
is required, simply because __libc_tls_initialized would always be set
to true inside libc.so.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-25-bugaevc@gmail.com>
Now that the signal code no longer accesses it, the only real user of it
was mig-reply.c, so move the logic for managing the port there.
If we're in SHARED and outside of rtld, we know that __LIBC_NO_TLS ()
always evaluates to 0, and a TLS reply port will always be used, not
__hurd_reply_port0. Still, the compiler does not see that
__hurd_reply_port0 is never used due to its address being taken. To deal
with this, explicitly compile out __hurd_reply_port0 when we know we
won't use it.
Also, instead of accessing the port via THREAD_SELF->reply_port, this
uses THREAD_GETMEM and THREAD_SETMEM directly, avoiding possible
miscompilations.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
This reverts commit b37899d34d.
Apparently we load libc.so (and thus start using its functions) before
calling TLS_INIT_TP, so libc.so functions should not actually assume
that TLS is always set up.
Previously, once we set up TLS, we would implicitly switch from using
__hurd_reply_port0 to reply_port inside the TCB, leaving the former
unused. But we never deallocated it, so it got leaked.
Instead, migrate the port into the new TCB's reply_port slot. This
avoids both the port leak and an extra syscall to create a new reply
port for the TCB.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-28-bugaevc@gmail.com>
If we're doing signals, that means we've already got the signal thread
running, and that implies TLS having been set up. So we know that
__hurd_local_reply_port will resolve to THREAD_SELF->reply_port, and can
access that directly using the THREAD_GETMEM and THREAD_SETMEM macros.
This avoids potential miscompilations, and should also be a tiny bit
faster.
Also, use mach_port_mod_refs () and not mach_port_destroy () to destroy
the receive right. mach_port_destroy () should *never* be used on
mach_task_self (); this can easily lead to port use-after-free
vulnerabilities if the task has any other references to the same port.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-26-bugaevc@gmail.com>
When glibc is built as a shared library, TLS is always initialized by
the call of TLS_INIT_TP () macro made inside the dynamic loader, prior
to running the main program (see dl-call_tls_init_tp.h). We can take
advantage of this: we know for sure that __LIBC_NO_TLS () will evaluate
to 0 in all other cases, so let the compiler know that explicitly too.
Also, only define _hurd_tls_init () and TLS_INIT_TP () under the same
conditions (either !SHARED or inside rtld), to statically assert that
this is the case.
Other than a microoptimization, this also helps with avoiding awkward
sharing of the __libc_tls_initialized variable between ld.so and libc.so
that we would have to do otherwise -- we know for sure that no sharing
is required, simply because __libc_tls_initialized would always be set
to true inside libc.so.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-25-bugaevc@gmail.com>
These are just regular local variables that are not accessed in any
funny ways, not even though a pointer. There's absolutely no reason to
declare them volatile. It only ends up hurting the quality of the
generated machine code.
If anything, it would make sense to decalre sigsp as *pointing* to
volatile memory (volatile void *sigsp), but evidently that's not needed
either.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230403115621.258636-2-bugaevc@gmail.com>
This is based on the Linux port's version, but laid out to match Mach's
struct i386_thread_state, much like the i386 version does.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Rename x86_cpu_INDEX_7_ECX_1 to x86_cpu_INDEX_7_ECX_15 for the unused bit
15 in ECX from CPUID with EAX == 0x7 and ECX == 0.
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
sysdeps/mach/hurd/htl/pt-pthread_self.c: New file.
htl/Makefile: .. Add it to libc routine.
sysdeps/mach/hurd/htl/pt-sysdep.c(__pthread_self): Remove it.
sysdeps/mach/hurd/htl/pt-sysdep.h(__pthread_self): Add hidden propertie.
htl/Versions(__pthread_self) Version it as private symbol.
Signed-off-by: Guy-Fleury Iteriteka <gfleury@disroot.org>
Message-Id: <20230318095826.1125734-3-gfleury@disroot.org>
As indicated by sparc kernel-features.h, even though sparc64 defines
__NR_pause, it is not supported (ENOSYS). Always use ppoll or the
64 bit time_t variant instead.
The error handling is moved to sysdeps/ieee754 version with no SVID
support. The compatibility symbol versions still use the wrapper
with SVID error handling around the new code. There is no new symbol
version nor compatibility code on !LIBM_SVID_COMPAT targets
(e.g. riscv).
The ia64 is unchanged, since it still uses the arch specific
__libm_error_region on its implementation. For both i686 and m68k,
which provive arch specific implementation, wrappers are added so
no new symbol are added (which would require to change the
implementations).
It shows an small improvement, the results for fmod:
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
x86_64 (Ryzen 9) | subnormals | 12.5049 | 9.40992
x86_64 (Ryzen 9) | normal | 296.939 | 296.738
x86_64 (Ryzen 9) | close-exponents | 16.0244 | 13.119
aarch64 (N1) | subnormal | 6.81778 | 4.33313
aarch64 (N1) | normal | 155.620 | 152.915
aarch64 (N1) | close-exponents | 8.21306 | 5.76138
armhf (N1) | subnormal | 15.1083 | 14.5746
armhf (N1) | normal | 244.833 | 241.738
armhf (N1) | close-exponents | 21.8182 | 22.457
Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
This uses a new algorithm similar to already proposed earlier [1].
With x = mx * 2^ex and y = my * 2^ey (mx, my, ex, ey being integers),
the simplest implementation is:
mx * 2^ex == 2 * mx * 2^(ex - 1)
while (ex > ey)
{
mx *= 2;
--ex;
mx %= my;
}
With mx/my being mantissa of double floating pointer, on each step the
argument reduction can be improved 8 (which is sizeof of uint32_t minus
MANTISSA_WIDTH plus the signal bit):
while (ex > ey)
{
mx << 8;
ex -= 8;
mx %= my;
} */
The implementation uses builtin clz and ctz, along with shifts to
convert hx/hy back to doubles. Different than the original patch,
this path assume modulo/divide operation is slow, so use multiplication
with invert values.
I see the following performance improvements using fmod benchtests
(result only show the 'mean' result):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
x86_64 (Ryzen 9) | subnormals | 17.2549 | 12.0318
x86_64 (Ryzen 9) | normal | 85.4096 | 49.9641
x86_64 (Ryzen 9) | close-exponents | 19.1072 | 15.8224
aarch64 (N1) | subnormal | 10.2182 | 6.81778
aarch64 (N1) | normal | 60.0616 | 20.3667
aarch64 (N1) | close-exponents | 11.5256 | 8.39685
I also see similar improvements on arm-linux-gnueabihf when running on
the N1 aarch64 chips, where it a lot of soft-fp implementation (for
modulo, and multiplication):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
armhf (N1) | subnormal | 11.6662 | 10.8955
armhf (N1) | normal | 69.2759 | 34.1524
armhf (N1) | close-exponents | 13.6472 | 18.2131
Instead of using the math_private.h definitions, I used the
math_config.h instead which is used on newer math implementations.
Co-authored-by: kirill <kirill.okhotnikov@gmail.com>
[1] https://sourceware.org/pipermail/libc-alpha/2020-November/119794.html
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
This uses a new algorithm similar to already proposed earlier [1].
With x = mx * 2^ex and y = my * 2^ey (mx, my, ex, ey being integers),
the simplest implementation is:
mx * 2^ex == 2 * mx * 2^(ex - 1)
while (ex > ey)
{
mx *= 2;
--ex;
mx %= my;
}
With mx/my being mantissa of double floating pointer, on each step the
argument reduction can be improved 11 (which is sizeo of uint64_t minus
MANTISSA_WIDTH plus the signal bit):
while (ex > ey)
{
mx << 11;
ex -= 11;
mx %= my;
} */
The implementation uses builtin clz and ctz, along with shifts to
convert hx/hy back to doubles. Different than the original patch,
this path assume modulo/divide operation is slow, so use multiplication
with invert values.
I see the following performance improvements using fmod benchtests
(result only show the 'mean' result):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
x86_64 (Ryzen 9) | subnormals | 19.1584 | 12.5049
x86_64 (Ryzen 9) | normal | 1016.51 | 296.939
x86_64 (Ryzen 9) | close-exponents | 18.4428 | 16.0244
aarch64 (N1) | subnormal | 11.153 | 6.81778
aarch64 (N1) | normal | 528.649 | 155.62
aarch64 (N1) | close-exponents | 11.4517 | 8.21306
I also see similar improvements on arm-linux-gnueabihf when running on
the N1 aarch64 chips, where it a lot of soft-fp implementation (for
modulo, clz, ctz, and multiplication):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
armhf (N1) | subnormal | 15.908 | 15.1083
armhf (N1) | normal | 837.525 | 244.833
armhf (N1) | close-exponents | 16.2111 | 21.8182
Instead of using the math_private.h definitions, I used the
math_config.h instead which is used on newer math implementations.
Co-authored-by: kirill <kirill.okhotnikov@gmail.com>
[1] https://sourceware.org/pipermail/libc-alpha/2020-November/119794.html
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
Linux kernel uses AT_HWCAP2 to indicate if FSGSBASE instructions are
enabled. If the HWCAP2_FSGSBASE bit in AT_HWCAP2 is set, FSGSBASE
instructions can be used in user space. Define dl_check_hwcap2 to set
the FSGSBASE feature to active on Linux when the HWCAP2_FSGSBASE bit is
set.
Add a test to verify that FSGSBASE is active on current kernels.
NB: This test will fail if the kernel doesn't set the HWCAP2_FSGSBASE
bit in AT_HWCAP2 while fsgsbase shows up in /proc/cpuinfo.
Reviewed-by: Florian Weimer <fweimer@redhat.com>
The divss instruction clobbers its first argument, and the constraints
need to reflect that. Fortunately, with GCC 12, generated code does
not actually change, so there is no externally visible bug.
Suggested-by: Jakub Jelinek <jakub@redhat.com>
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Just like the other existing rtld-str* files, this provides rtld with
usable versions of stpncpy and strncpy.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-22-bugaevc@gmail.com>
The source code is the same as sysdeps/i386/htl/tcb-offsets.sym, but of
course the produced tcb-offsets.h will be different.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-21-bugaevc@gmail.com>
These do not need any changes to be used on x86_64.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-20-bugaevc@gmail.com>
This is more correct, if only because these fields are defined as having
the type unsigned int in the Mach headers, so casting them to a signed
int and then back is suboptimal.
Also, remove an extra reassignment of uesp -- this is another remnant of
the ecx kludge.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-16-bugaevc@gmail.com>
There's nothing Mach- or Hurd-specific about it; any port that ends
up with rtld pulling in strncpy will need this.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-15-bugaevc@gmail.com>
This was used for the value of libc-lock's owner when TLS is not yet set
up, so THREAD_SELF can not be used. Since the value need not be anything
specific -- it just has to be non-NULL -- we can just use a plain
constant, such as (void *) 1, for this. This avoids accessing the symbol
through GOT, and exporting it from libc.so in the first place.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-12-bugaevc@gmail.com>
Noone is or should be using __hurd_threadvar_stack_{offset,mask}, we
have proper TLS now. These two remaining variables are never set to
anything other than zero, so any code that would try to use them as
described would just dereference a zero pointer and crash. So remove
them entirely.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230319151017.531737-6-bugaevc@gmail.com>
And make always supported. The configure option was added on glibc 2.25
and some features require it (such as hwcap mask, huge pages support, and
lock elisition tuning). It also simplifies the build permutations.
Changes from v1:
* Remove glibc.rtld.dynamic_sort changes, it is orthogonal and needs
more discussion.
* Cleanup more code.
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
Prevent sh from interpreting a user string as shell options if it
starts with '-' or '+'. Since the version of /bin/sh used for testing
system() is different from the full-fledged system /bin/sh add support
to it for handling "--" after "-c". Add a testcase to ensure the
expected behavior.
Signed-off-by: Joe Simmons-Talbott <josimmon@redhat.com>
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Instead define the required fields in system dependend files. The only
system dependent definition is FILENAME_MAX, which should match POSIX
PATH_MAX, and it is obtained from either kernel UAPI or mach headers.
Currently set pre-defined value from current kernels.
It avoids a circular dependendy when including stdio.h in
gen-as-const-headers files.
Checked on x86_64-linux-gnu and i686-linux-gnu
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
They are both used by __libc_freeres to free all library malloc
allocated resources to help tooling like mtrace or valgrind with
memory leak tracking.
The current scheme uses assembly markers and linker script entries
to consolidate the free routine function pointers in the RELRO segment
and to be freed buffers in BSS.
This patch changes it to use specific free functions for
libc_freeres_ptrs buffers and call the function pointer array directly
with call_function_static_weak.
It allows the removal of both the internal macros and the linker
script sections.
Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
This allows other targets to use the same inputs for their own libmvec
microbenchmarks without having to duplicate them in their own
subdirectory.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Binutils 2.40 sets EF_LARCH_OBJABI_V1 for shared objects:
$ ld --version | head -n1
GNU ld (GNU Binutils) 2.40
$ echo 'int dummy;' > dummy.c
$ cc dummy.c -shared
$ readelf -h a.out | grep Flags
Flags: 0x43, DOUBLE-FLOAT, OBJ-v1
We need to ignore it in ldconfig or ldconfig will consider all shared
objects linked by Binutils 2.40 "unsupported". Maybe we should stop
setting EF_LARCH_OBJABI_V1 for shared objects, but Binutils 2.40 is
already released and we cannot change it.
Linux threads were removed about 12 years ago and the current
nptl implementation only requires 4-byte alignment for pthread
locks.
The 16-byte alignment causes various issues. For example in
building ignition-msgs, we have:
/usr/include/google/protobuf/map.h:124:37: error: static assertion failed
124 | static_assert(alignof(value_type) <= 8, "");
| ~~~~~~~~~~~~~~~~~~~~^~~~
This is caused by the 16-byte pthread lock alignment.
Signed-off-by: John David Anglin <dave.anglin@bell.net>
For better debug experience use separate code block with extra
cfi_* directives to run child (same as in __clone3).
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Use the clone3 wrapper on ARC. It doesn't care about stack alignment.
All callers should provide an aligned stack.
It follows the internal signature:
extern int clone3 (struct clone_args *__cl_args, size_t __size,
int (*__func) (void *__arg), void *__arg);
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Between versions v2.11 and v2.12 struct ntptimeval got new fields.
That wasn't a problem because new function ntp_gettimex was created
(and made default) to support new struct. Old ntp_gettime was not
using new fields so it was safe to call with old struct
definition. Then commits 5613afe9e3 and b6ad64b907 (added for
64 bit time_t support), ntp_gettime start setting new fields.
Sets fields manually to maintain compatibility with v2.11 struct
definition.
Resolves#30156
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
according to man-pages-posix-2017, shm_open() function may fail if the length
of the name argument exceeds {_POSIX_PATH_MAX} and set ENAMETOOLONG
Signed-off-by: abushwang <abushwangs@gmail.com>
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Recorded in [BZ #30183]:
1. export GLIBC_TUNABLES=glibc.cpu.hwcaps=-AVX512
2. Add _dl_printf("p -- %s\n", p); just before switch(nl) in
sysdeps/x86/cpu-tunables.c
3. compiled and run ./testrun.sh /usr/bin/ls
you will get:
p -- -AVX512
p -- LC_ADDRESS=en_US.UTF-8
p -- LC_NUMERIC=C
...
The function, TUNABLE_CALLBACK (set_hwcaps)
(tunable_val_t *valp), checks far more than it should and it
should stop at end of "-AVX512".
Fix bug that SIGCHLD is erroneously blocked forever in the following
scenario:
1. Thread A calls system but hasn't returned yet
2. Thread B calls another system but returns
SIGCHLD would be blocked forever in thread B after its system() returns,
even after the system() in thread A returns.
Although POSIX does not require, glibc system implementation aims to be
thread and cancellation safe. This bug was introduced in
5fb7fc9635 when we moved reverting signal
mask to happen when the last concurrently running system returns,
despite that signal mask is per thread. This commit reverts this logic
and adds a test.
Signed-off-by: Adam Yi <ayi@janestreet.com>
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
This patch updates the kernel version in the tests tst-mman-consts.py,
tst-mount-consts.py and tst-pidfd-consts.py to 6.2. (There are no new
constants covered by these tests in 6.2 that need any other header
changes, and the removed MAP_VARIABLE for hppa was addressed
separately.)
Tested with build-many-glibcs.py.
The __builtin_arm_uqsub8 is an internal GCC builtin which might change
in future release (the correct way is to include "arm_acle.h" and use
__uqsub8 ()). Since not all compilers support it, just use the
inline assembler instead.
Checked on armv7a-linux-gnueabihf.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
The generic implementation already cover word access along with
cmpbge for both aligned and unaligned, so use it instead.
Checked qemu static for alpha-linux-gnu.
The default, and power7 implementation just adds word aligned
access when inputs have the same aligment. The unaligned case
is still done by byte operations.
This is already covered by the generic implementation, which also add
the unaligned input optimization.
Checked on powerpc64-linux-gnu built without multi-arch for powerpc64,
power7, power8, and power9 (build for le).
Reviewed-by: Rajalakshmi Srinivasaraghavan <rajis@linux.ibm.com>
The default, power4, and power7 implementation just adds word aligned
access when inputs have the same aligment. The unaligned case
is still done by byte operations.
This is already covered by the generic implementation, which also add
the unaligned input optimization.
Checked on powerpc-linux-gnu built without multi-arch for powerpc,
power4, and power7.
Reviewed-by: Rajalakshmi Srinivasaraghavan <rajis@linux.ibm.com>
C2x adds binary integer constants starting with 0b or 0B, and supports
those constants for the %i scanf format (in addition to the %b format,
which isn't yet implemented for scanf in glibc). Implement that scanf
support for glibc.
As with the strtol support, this is incompatible with previous C
standard versions, in that such an input string starting with 0b or 0B
was previously required to be parsed as 0 (with the rest of the input
potentially matching subsequent parts of the scanf format string).
Thus this patch adds 12 new __isoc23_* functions per long double
format (12, 24 or 36 depending on how many long double formats the
glibc configuration supports), with appropriate header redirection
support (generally very closely following that for the __isoc99_*
scanf functions - note that __GLIBC_USE (DEPRECATED_SCANF) takes
precedence over __GLIBC_USE (C2X_STRTOL), so the case of GNU
extensions to C89 continues to get old-style GNU %a and does not get
this new feature). The function names would remain as __isoc23_* even
if C2x ends up published in 2024 rather than 2023.
When scanf %b support is added, I think it will be appropriate for all
versions of scanf to follow C2x rules for inputs to the %b format
(given that there are no compatibility concerns for a new format).
Tested for x86_64 (full glibc testsuite). The first version was also
tested for powerpc (32-bit) and powerpc64le (stdio-common/ and wcsmbs/
tests), and with build-many-glibcs.py.
Before GCC r13-2728, it would produce a normal dynamic-linked executable
with -static-pie. I mistakely believed it would produce a static-linked
executable, so failed to detect the breakage. Then with Binutils 2.40
and (vanilla) GCC 12, libc_cv_static_pie_on_loongarch is mistakenly
enabled and cause a building failure with "undefined reference to
_DYNAMIC".
Fix the issue by disabling static PIE if -static-pie creates something
with a INTERP header.
"We don't need it any more"
The INTR_MSG_TRAP macro in intr-msg.h used to play little trick with
the stack pointer: it would temporarily save the "real" stack pointer
into ecx, while setting esp to point to just before the message buffer,
and then invoke the mach_msg trap. This way, INTR_MSG_TRAP reused the
on-stack arguments laid out for the containing call of
_hurd_intr_rpc_mach_msg (), passing them to the mach_msg trap directly.
This, however, required special support in hurdsig.c and trampoline.c,
since they now had to recognize when a thread is inside the piece of
code where esp doesn't point to the real tip of the stack, and handle
this situation specially.
Commit 1d20f33ff4 has removed the actual
temporary change of esp by actually re-pushing mach_msg arguments onto
the stack, and popping them back at end. It did not, however, deal with
the rest of "the ecx kludge" code in other files, resulting in potential
crashes if a signal arrives in the middle of pushing arguments onto the
stack.
Fix that by removing "the ecx kludge". Instead, when we want a thread
to skip the RPC, but cannot make just make it jump to after the trap
since it's not done adjusting the stack yet, set the SYSRETURN register
to MACH_SEND_INTERRUPTED (as we do anyway), and rely on the thread
itself for detecting this case and skipping the RPC.
This simplifies things somewhat and paves the way for a future x86_64
port of this code.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230301162355.426887-1-bugaevc@gmail.com>
The _FPU_SETCW and _FPU_GETCW macros are defined with inline assemblies.
They use the sfpc and efpc instructions, respectively. But both contain
a spurious second operand that leads to a compile error with Clang.
Removing this operand works both with gcc/gas (since binutils 2.18) as
well as with clang/llvm.
Needed due to recent commits:
- "added pair of inputs for hypotf in binary32"
commit ID cf7ffdd8a5
- "update auto-libm-test-out-hypot"
commit ID 3efbf11fdf
Linux 6.2 adds six new Arm HWCAP values and two new HWCAP2 values; add
them to glibc's Arm bits/hwcap.h, with corresponding dl-procinfo.c and
dl-procinfo.h updates.
Tested with build-many-glibcs.py for arm-linux-gnueabi.
This is for future-proofing. On i386, it is 4-byte aligned anyway, but
on x86_64, we want it 8-byte aligned, not 4-byte aligned.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Message-Id: <20230214173722.428140-4-bugaevc@gmail.com>
Update libm test ulps for
commit 3efbf11fdf
Author: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Date: Tue Feb 14 11:24:59 2023 +0100
update auto-libm-test-out-hypot
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
This patch implements the LoongArch specific math barriers in order to omit
the store and load from stack if possible.
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The Linux kernel upstream commit 71bdea6f798b ("parisc: Align parisc
MADV_XXX constants with all other architectures") dropped the
parisc-specific MADV_* values in favour of the same constants as
other architectures. In the same commit a wrapper was added which
translates the old values to the standard MADV_* values to avoid
breakage of existing programs.
This upstream patch has been downported to all stable kernel trees as
well.
This patch now drops the parisc specific constants from glibc to
allow newly compliled programs to use the standard MADV_* constants.
v2: Added NEWS section, based on feedback from Florian Weimer
Signed-off-by: Helge Deller <deller@gmx.de>
This drops all of the return address rewriting kludges. The only
remaining hack is the jump out of a call stack while adjusting the
stack pointer.
Signed-off-by: Sergey Bugaev <bugaevc@gmail.com>
Linux 6.2 has no new syscalls. Update the version number in
syscall-names.list to reflect that it is still current for 6.2.
Tested with build-many-glibcs.py.
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.
Linux 6.2 removed the hppa compatibility MAP_VARIABLE define. That
means that, whether or not we remove it in glibc, it needs to be
ignored in tst-mman-consts.py (since this macro comparison
infrastructure expects that new kernel header versions only add new
macros, not remove old ones).
Tested with build-many-glibcs.py for hppa-linux-gnu (Linux 6.2
headers).
Fix the computation to allow for cntfrq_el0 being larger than 1GHz.
Assume cntfrq_el0 is a multiple of 1MHz to increase the maximum
interval (1024 seconds at 1GHz).
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
It fixes the build after 7ea510127e and 22999b2f0f.
Checked with build for s390x-linux-gnu with -march=z13.
Reviewed-by: Arjun Shankar <arjun@redhat.com>
__builtin_arm_uqsub8 is only available on gcc newer or equal than 10.
Checked on arm-linux-gnueabihf built with gcc 9.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
The default Linux implementation already handled the Linux generic
ABIs interface used on newer architectures, so there is no need to
Imply the generic any longer.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Tested-by: Carlos O'Donell <carlos@redhat.com>
