glibc/sysdeps/mach/hurd/Makefile
Jeremie Koenig 653d74f12a hurd: Global signal disposition
This adds _hurd_sigstate_set_global_rcv used by libpthread to enable
POSIX-confirming behavior of signals on a per-thread basis.

This also provides a sigstate destructor _hurd_sigstate_delete, and a
global process signal state, which needs to be locked and check when
global disposition is enabled, thus the addition of _hurd_sigstate_lock
_hurd_sigstate_actions _hurd_sigstate_pending _hurd_sigstate_unlock helpers.

This also updates all the glibc code accordingly.

This also drops support for get_int(INIT_SIGMASK), which did not make sense
any more since we do not have a single signal thread any more.

During fork/spawn, this also reinitializes the child global sigstate's
lock. That cures an issue that would very rarely cause a deadlock in the
child in fork, tries to unlock ss' critical section lock at the end of
fork.  This will typically (always?) be observed in /bin/sh, which is not
surprising as that is the foremost caller of fork.

To reproduce an intermediate state, add an endless loop if
_hurd_global_sigstate is locked after __proc_dostop (cast through
volatile); that is, while still being in the fork's parent process.

When that triggers (use the libtool testsuite), the signal thread has
already locked ss (which is _hurd_global_sigstate), and is stuck at
hurdsig.c:685 in post_signal, trying to lock _hurd_siglock (which the
main thread already has locked and keeps locked until after
__task_create).  This is the case that ss->thread == MACH_PORT_NULL, that
is, a global signal.  In the main thread, between __proc_dostop and
__task_create is the __thread_abort call on the signal thread which would
abort any current kernel operation (but leave ss locked).  Later in fork,
in the parent, when _hurd_siglock is unlocked in fork, the parent's
signal thread can proceed and will unlock eventually the global sigstate.
In the client, _hurd_siglock will likewise be unlocked, but the global
sigstate never will be, as the client's signal thread has been configured
to restart execution from _hurd_msgport_receive.  Thus, when the child
tries to unlock ss' critical section lock at the end of fork, it will
first lock the global sigstate, will spin trying to lock it, which can
never be successful, and we get our deadlock.

Options seem to be:

  * Move the locking of _hurd_siglock earlier in post_signal -- but that
    may generally impact performance, if this locking isn't generally
    needed anyway?

    On the other hand, would it actually make sense to wait here until we
    are not any longer in a critical section (which is meant to disable
    signal delivery anyway (but not for preempted signals?))?

  * Clear the global sigstate in the fork's child with the rationale that
    we're anyway restarting the signal thread from a clean state.  This
    has now been implemented.

Why has this problem not been observed before Jérémie's patches?  (Or has
it?  Perhaps even more rarely?)  In _S_msg_sig_post, the signal is now
posted to a *global receiver thread*, whereas previously it was posted to
the *designated signal-receiving thread*.  The latter one was in a
critical section in fork, so didn't try to handle the signal until after
leaving the critical section?  (Not completely analyzed and verified.)

Another question is what the signal is that is being received
during/around the time __proc_dostop executes.
2019-12-29 18:32:49 +01:00

217 lines
7.8 KiB
Makefile
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Copyright (C) 1993-2019 Free Software Foundation, Inc.
# This file is part of the GNU C Library.
# The GNU C Library is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
# License as published by the Free Software Foundation; either
# version 2.1 of the License, or (at your option) any later version.
# The GNU C Library is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
# Lesser General Public License for more details.
# You should have received a copy of the GNU Lesser General Public
# License along with the GNU C Library; if not, see
# <https://www.gnu.org/licenses/>.
ifdef in-Makerules
# Look for header files in hurd/ under the top-level library source directory.
# Look for generated header files where they get created.
includes += -I$(..)hurd -I$(common-objpfx)hurd/
# We use the style `if (err = call(...))' a lot in the Hurd code,
# where we have a lot of functions that return zero or an errno code.
+cflags += -Wno-parentheses
# Do not use any assembly code from sysdeps/unix (and subdirectories).
# This bypasses all the system call stubs and uses any existing posix or
# generic C files instead.
inhibit-sysdep-asm += unix.*
inhibit-unix-syscalls = yes
# Don't try to generate anything from the installed Unix system and its
# libraries. That is only of use when building for a Unix system, so as to
# be compatible with some existing binaries for that system.
inhibit-glue = yes
ifeq (,$(filter mach hurd,$(subdir)))
# Subdirectories other than hurd/ might use the generated Hurd headers.
# So make sure we get a chance to run in hurd/ to make them before all else.
# (But we don't want to do this in mach/, because hurd/ needs some things
# there, and we know mach/ doesn't need anything from hurd/.)
hurd-objpfx = $(common-objpfx)hurd/
# These are all the generated headers that <hurd.h> includes.
before-compile += $(patsubst %,$(hurd-objpfx)hurd/%.h,auth io fs process)
$(patsubst %,$(hurd-objpfx)hurd/%.h,auth io fs process): hurd-before-compile
.PHONY: hurd-before-compile
hurd-before-compile: $(common-objpfx)mach/mach-shortcuts.h
$(MAKE) -C $(..)hurd subdir=hurd before-compile no_deps=t
endif
# Hurd profil.c includes this file, so give a rule to make it.
ifeq ($(subdir),gmon)
$(common-objpfx)hurd/../mach/RPC_task_get_sampled_pcs.c:
$(MAKE) -C $(..)mach subdir=mach before-compile no_deps=t
endif
# Generate bits/errno.h from the section of the manual that lists all the errno
# codes.
errno.texinfo = $(..)manual/errno.texi
hurd = $(..)sysdeps/mach/hurd
define mach-errno-h
($(foreach h,mach/message.h \
mach/kern_return.h \
mach/mig_errors.h \
device/device_types.h,\
echo '#include <$h>';\
))
endef
# We use the compiler to generate a list of absolute file names for
# the headers we want to search for Mach error codes, listed above (and
# incidentally, all other headers those include).
-include $(common-objpfx)errnos.d
$(common-objpfx)errnos.d: $(mach-errnos-deps)
$(mach-errno-h) | \
$(CC) $(CFLAGS) \
$(subst -include $(common-objpfx)libc-modules.h,,$(CPPFLAGS)) \
-M -x c - | \
sed $(sed-remove-objpfx) -e 's,- *:,mach-errnos-deps :=,' \
-e 's, \.\./, $(..),g' > $@t
mv -f $@t $@
$(hurd)/bits/errno.h: $(common-objpfx)stamp-errnos ;
$(common-objpfx)stamp-errnos: $(hurd)/errnos.awk $(errno.texinfo) \
$(mach-errnos-deps) $(common-objpfx)errnos.d
mkdir -p $(hurd-objpfx)bits
$(AWK) -f $^ > $(hurd-objpfx)bits/errno.h-tmp
# Make it unwritable so noone will edit it by mistake.
-chmod a-w $(hurd-objpfx)bits/errno.h-tmp
$(move-if-change) $(hurd-objpfx)bits/errno.h-tmp $(hurd)/bits/errno.h
touch $@
common-generated += errnos.d stamp-errnos
# We install the real libc.a as libcrt.a and as libc.a we install a linker
# script which does -( -lcrt -lmachuser -lhurduser -).
libc-name = crt
ifeq (,$(subdir))
install-others += $(inst_libdir)/libc.a
$(inst_libdir)/libc.a: $(hurd)/libc-ldscript $(+force); $(do-install)
ifeq (yes,$(build-profile))
install-others += $(inst_libdir)/libc_p.a
$(inst_libdir)/libc_p.a: $(hurd)/libc_p-ldscript $(+force); $(do-install)
endif
endif
# Make sure these are used to build the libc.so shared object too. There
# is a circular dependency between each of these shared objects and libc
# (many high-level libc functions call stubs, stubs call low-level libc
# functions like memcpy and mach_msg). This works out fine at run time
# (all the objects are loaded before resolving their symbols, so these
# interdependencies are fine). But to create the shared objects we must
# link them one at a time; since each needs one or both of the others to
# produce its DT_NEEDED entries and to assign its undefined symbols the
# right symbol versions, we can't do any of them before the others! To
# get around this, we link each lib*user.so shared object twice. First,
# we link an object without reference to libc.so (since we haven't linked
# libc.so yet), so it lacks a DT_NEEDED record for the libc soname it
# depends on, and its undefined symbol references lack the symbol version
# assignments they should have. We will use this shared object solely to
# link libc.so against it; that gives libc.so the proper DT_NEEDED record,
# and symbol versions assignments (if the lib*user.so object is using them).
# Finally we link a second version of the same lib*user.so shared object,
# this time linked normally against libc so it gets a proper DT_NEEDED
# record and symbol version set; this one can be installed for run-time use.
rpcuserlibs := $(common-objpfx)mach/libmachuser.so \
$(common-objpfx)hurd/libhurduser.so
link-rpcuserlibs := $(rpcuserlibs:%user.so=%user-link.so)
$(common-objpfx)libc.so: $(link-rpcuserlibs)
$(common-objpfx)linkobj/libc.so: $(link-rpcuserlibs)
rpath-dirs += mach hurd
# Make sure the `lib' pass builds the dummy shared objects so
# we can link libc against them.
ifeq (mach,$(subdir))
lib-noranlib: $(common-objpfx)mach/libmachuser-link.so
endif
ifeq (hurd,$(subdir))
lib-noranlib: $(common-objpfx)hurd/libhurduser-link.so
endif
$(link-rpcuserlibs): %-link.so: %_pic.a
# These shared objects are just for the purpose of linking libc,
# so they don't need abi-note.o linked into them.
$(build-shlib-helper) \
-nostdlib -o $@ \
-Wl,-soname=$(*F).so$($(*F).so-version) \
$(build-shlib-objlist)
libmachuser-link.so-no-z-defs = yes
libhurduser-link.so-no-z-defs = yes
# And get them into the libc.so ldscript.
$(inst_libdir)/libc.so: $(rpcuserlibs)
# The RPC stubs from these libraries are needed in building the dynamic
# linker, too. It must be self-contained, so we link the needed PIC
# objects directly into the shared object.
ifeq (elf,$(subdir))
$(objpfx)librtld.map: $(rpcuserlibs:.so=_pic.a)
CFLAGS-dl-load.c = -DEXTERNAL_MAP_FROM_FD
endif
# Override the generic Makeconfig values so we link against the RPC libs.
link-libc-static := -Wl,--start-group \
$(patsubst %,$(common-objpfx)%.a,\
libc mach/libmachuser hurd/libhurduser) \
$(static-gnulib) -Wl,--end-group
link-libc-static-tests := -Wl,--start-group \
$(patsubst %,$(common-objpfx)%.a,\
libc mach/libmachuser hurd/libhurduser) \
$(static-gnulib-tests) -Wl,--end-group
ifeq ($(subdir),csu)
extra-objs += static-start.o
# We need special startup code for statically linked binaries.
$(objpfx)crt0.o: $(objpfx)static-start.o $(objpfx)abi-note.o $(objpfx)init.o
$(link-relocatable)
endif
ifeq (hurd, $(subdir))
sysdep_routines += cthreads
endif
ifeq (io, $(subdir))
sysdep_routines += f_setlk close_nocancel_nostatus
endif
ifeq ($(subdir),sunrpc)
sysdep_headers += nfs/nfs.h
endif
ifeq ($(subdir),socket)
sysdep_headers += net/ethernet.h net/if_arp.h net/if_ether.h net/route.h
endif
ifeq ($(subdir),nis)
CFLAGS-ypclnt.c += -DUSE_BINDINGDIR=1
endif
LDLIBS-pthread.so += $(objdir)/hurd/libhurduser.so
endif # in-Makerules