Safely recognize when fork is running from main thread or another pthread
* child_info.h (struct child_info): Add member from_main.
* fork.cc (frok::child): Check from_main rather than stackaddr.
(frok::parent): Set ch.from_main if running in the main thread.
Rename parent stack members in child_info struct to align with OS names
* child_info.h (CURR_CHILD_INFO_MAGIC): Align to below change.
(class child_info_fork): Rename stacktop to stacklimit. Rename
stackbottom to stackbase. Accommodate name change throughout Cygwin.
Rephrase comments to be clearer.
Drop using _tlsbase and _tlstop in favor of access via NtCurrentTeb.
* cygtls.h (_tlsbase): Remove. Replace throughout with
NtCurrentTeb()->Tib.StackBase.
(_tlstop): Remove. Replace throughout with
NtCurrentTeb()->Tib.StackLimit.
* dcrt0.cc (child_info_fork::alloc_stack): Move definition of local
teb variable up to be used throughout.
* include/cygwin/config.h (__getreent): Use inline function on both
architectures.
Corinna Vinschen [Fri, 27 Nov 2015 13:39:11 +0000 (14:39 +0100)]
Fix race condition when waiting for a signal
* cygtls.h (_cygtls::wait_signal_arrived): Renamed from
set_signal_arrived.
(_cygtls::set_signal_arrived): New function signalling signal_arrived.
(_cygtls::reset_signal_arrived): Don't reset will_wait_for_signal.
(_cygtls::unwait_signal_arrived): New function only resetting
will_wait_for_signal.
(class wait_signal_arrived): Rename from set_signal_arrived.
Accommodate name change throughout Cygwin.
(wait_signal_arrived::~wait_signal_arrived): Call
_cygtls::unwait_signal_arrived. Add comment.
* cygserver_ipc.h (ipc_set_proc_info): Fetch signal_arrived handle
via call to _cygtls::get_signal_arrived.
* exceptions.cc (_cygtls::interrupt_setup): Signal signal_arrived via
call to _cygtls::set_signal_arrived.
(_cygtls::handle_SIGCONT): Ditto.
* fhandler_socket.cc (fhandler_socket::wait_for_events): Generate
WSAEVENT array prior to entering wait loop. Add cancel event object
if available. Remove calls to pthread_testcancel and just call
pthread::static_cancel_self if the cancel event object is signalled.
David Macek [Wed, 25 Nov 2015 12:49:42 +0000 (13:49 +0100)]
Add MacType to BLODA
One more patch. MacType was observed by several users to cause `GPGME:
Invalid crypto engine` failures in MSYS2. See
<https://github.com/Alexpux/MSYS2-packages/issues/393>.
I also removed two full stops in the sake of consistency.
* faq-using.xml: Add MacType to the BLODA. Fix formatting.
David Macek [Tue, 24 Nov 2015 22:29:34 +0000 (23:29 +0100)]
Add Forefront TMG to the BLODA
A user of MSYS2 had non-deterministic segfaults when running simple
command from bash started from cmd. We've isolated the cause to
`C:\Program Files (x86)\Forefront TMG Client\FwcWsp64.dll`. See
<http://sourceforge.net/p/msys2/tickets/212/> for a little more
information.
It seems enough to include the software in BLODA, but if not, sorry for
the unsolicited patch and let's discuss.
Marcus Shawcroft [Fri, 13 Nov 2015 13:53:38 +0000 (13:53 +0000)]
[ARM] Adjust strcmp-armv7.S selection logic.
Reformulate the strcmp-armv7.S selection logic around the architecture
features required by the implementation code rather (some) version of
the architecture that expose those features.
Corinna Vinschen [Sat, 21 Nov 2015 15:51:12 +0000 (16:51 +0100)]
Fix numeric and monetary decimal point and thousands separator in fa_IR and ps_AF locales
* nlsfuncs.cc (setlocaleinfo): New macro calling __setlocaleinfo.
(__setlocaleinfo): New function to set a locale-specific character
to an explicit wchar_t value.
(__set_lc_numeric_from_win): Handle fa_IR and ps_AF locales to return
same decimal point and thousands separator characters as on Linux.
(__set_lc_monetary_from_win): Ditto for monetary characters.
Corinna Vinschen [Tue, 10 Nov 2015 09:00:33 +0000 (10:00 +0100)]
get_posix_access: Skip merging permissions if user can't get more permissions
* sec_acl.cc (get_posix_access): Skip merging permissions if current
user has rwx permissions anywa, or if the sum of all group and other
permissions is less than or equal to the user's permissions.
Fix file type mode bit handling on object security.
* fhandler_tty.cc (fhandler_pty_slave::open): Add S_IFCHR flag
in call to create_object_sd_from_attribute.
(fhandler_pty_slave::fstat): Ditto in call to get_object_attribute.
(fhandler_pty_slave::fchmod): Add an orig_mode flag and ditto for
both calls.
(fhandler_pty_master::setup): Ditto in call to
create_object_sd_from_attribute.
* security.cc (get_object_attribute): Never add S_IFCHR to mode
here. Let the caller decide.
(create_object_sd_from_attribute): Ditto.
* setfacl.c (action_t): Rename DeleteAll to DeleteExt. Add
DeleteAll. Rearrange for bit-wise testing later in the code.
(delallacl): Handle -b -k combination.
(setfacl): Handle DeleteExt/DeleteAll.
(usage): Fix -b/-k output. Rearrange output to better fill 80
columns.
(main): Allow to combine -b and -k.
* security.h (authz_get_user_attribute): Declare bool.
* sec_helper.cc (authz_ctx::get_user_attribute): Make bool method.
Set S_IxOTH bits in returned attributes rather than S_IxUSR bits.
(authz_get_user_attribute): Make bool function.
* sec_acl.cc (get_posix_access): Introduce cygsid array to keep
track of all SIDs in the ACL. Move AuthZ calls into !new_style
permission post processing. When not using AuthZ, use
CheckTokenMembership to collect group permissions.
* fhandler_disk_file.cc (fhandler_disk_file::fchmod): Disable
deviation from POSIX 1003.1e in terms of GROUP_OBJ/CLASS_OBJ
permissions. Follow POSIX 1003.1e again. Keep old code in
for future reference.
* sec_acl.cc: Accommodate changes in ACE creation in leading
comment.
(set_posix_access): Fix user deny ACE creation. Split group
deny ACE creation into two steps, one to reflect CLASS_OBJ,
the other to reflect OTHER_OBJ.
Corinna Vinschen [Wed, 24 Jun 2015 09:52:29 +0000 (11:52 +0200)]
Reapply POSIX ACL changes.
- New, unified implementation of POSIX permission and ACL handling. The
new ACLs now store the POSIX ACL MASK/CLASS_OBJ permission mask, and
they allow to inherit the S_ISGID bit. ACL inheritance now really
works as desired, in a limited, but theoretically equivalent fashion
even for non-Cygwin processes.
To accommodate Windows default ACLs, the new code ignores SYSTEM and
Administrators group permissions when computing the MASK/CLASS_OBJ
permission mask on old ACLs, and it doesn't deny access to SYSTEM and
Administrators group based on the value of MASK/CLASS_OBJ when
creating the new ACLs.
The new code now handles the S_ISGID bit on directories as on Linux:
Setting S_ISGID on a directory causes new files and subdirs created
within to inherit its group, rather than the primary group of the user
who created the file. This only works for files and directories
created by Cygwin processes.
2015-05-29 Corinna Vinschen <corinna@vinschen.de>
Reapply POSIX ACL changes.
* utils.xml (setfacl): Show new option output.
(getfacl): Show new option output.
* sec_acl.cc (get_posix_access): Check for Cygwin "standard" ACL.
Apply umask, if so. Align comments.
* security.cc (set_created_file_access): Fix permission masking by
incoming requested file mode.
* sec_acl.cc (set_posix_access): Apply mask only in terms of execute bit
for SYSTEM and Admins group.
* sec_acl.cc (set_posix_access): Don't create DENY ACEs for USER and
GROUP entries if they are the same as USER_OBJ or GROUP_OBJ.
* fhandler.h (fhandler_pty_slave::facl): Add prototype.
* fhandler_tty.cc (fhandler_pty_slave::facl): New method.
(fhandler_pty_slave::fchown): Fix uid/gid handling.
* sec_acl.cc (set_posix_access): Drop superfluous class_idx variable.
Simplify and move around code in a few places. To improve ACL
readability, add r/w permissions to Admins ACE appended to pty ACL.
Add comment to explain Windows ACE Mask filtering being in the way of
creating a real CLASS_OBJ.
(get_posix_access): Fake CLASS_OBJ for ptys. Explain why.
* security.cc (get_object_attribute): Add S_IFCHR flag to attributes
when calling get_posix_access.
* sec_acl.cc (set_posix_access): Move merging group perms into owner
perms in case of owner == group after mask has been computed. Take
mask into account when doing so to avoid unnecessary ACCESS_DENIED_ACE.
* sec_acl.cc (get_posix_access): Only set saw_group_obj flag if we saw
the ACCESS_ALLOWED_ACE.
* fhandler_disk_file.cc (fhandler_disk_file::fchmod): Deliberatly
set GROUP_OBJ and CLASS_OBJ perms to new group perms. Add comment
to explain why.
* security.cc (set_created_file_access): Ditto.
* sec_acl.cc (set_posix_access): Replace previous patch. Return
EINVAL if uid and/or guid is invalid and not backed by an actual
Windows account.
* sec_acl.cc (set_posix_access): Workaround owner/group SIDs being NULL.
* sec_acl.cc (set_posix_access): Handle files with owner == group.
Rephrase switch statement checking against unfiltered a_type value.
(get_posix_access): Handle files with owner == group.
* sec_acl.cc (get_posix_access): Don't use GROUP_OBJ access to fix up
CLASS_OBJ mask on old-style ACLs. Fix a comment.
* sec_acl.cc (set_posix_access): Always make sure Admins have
WRITE_DAC and WRITE_OWNER permissions.
* security.h (create_object_sd_from_attribute): Drop handle parameter
from prototype.
* security.cc (create_object_sd_from_attribute): Drop handle parameter.
Just create the standard POSIXy security descriptor.
(set_object_attribute): Accommodate dropped paramter in call to
create_object_sd_from_attribute.
* fhandler_tty.cc: Ditto, throughout.
* fhandler_disk_file.cc (fhandler_disk_file::fchmod): Fix typo in
mask computation.
* fhandler.cc (fhandler_base::open_with_arch): Call open with mode
not umasked.
(fhandler_base::open): Explicitely umask mode on NFS here. Call new
set_created_file_access rather than set_file_attribute.
* fhandler_disk_file.cc (fhandler_disk_file::fchmod): Reimplement
setting permissions on filesystems supporting ACLs using the new
set_posix_access call.
(fhandler_disk_file::fchown): Ditto.
(fhandler_disk_file::mkdir): Call new set_created_file_access rather
than set_file_attribute.
* fhandler_socket.cc (fhandler_socket::bind): Don't umask here. Add
WRITE_OWNER access to allow writing group in case of SGID bit set.
Call new set_created_file_access rather than set_file_attribute.
* path.cc (symlink_worker): Call new set_created_file_access rather
than set_file_attribute.
* sec_acl.cc (searchace): Un-staticize.
(set_posix_access): New, complementary functionality to
get_posix_access.
(setacl): Implement in terms of get_posix_access/set_posix_access.
(get_posix_access): Add handling for just created files requiring
their first Cygwin ACL. Fix new_style recognition. Handle SGID
bit. For old-style ACLs, ignore SYSTEM and Administrators when
computing the {DEF_}CLASS_OBJ perms.
* security.cc (get_file_sd): Revamp comment. Change and (hopefully)
speed up inheritance processing for just created files.
(alloc_sd): Remove.
(set_security_attribute): Call set_posix_access instead of alloc_sd.
(get_object_attribute): Fix return value.
(create_object_sd_from_attribute): Call set_posix_access instead of
alloc_sd.
(set_file_attribute): Remove.
(set_created_file_access): New function implemented in terms of
get_posix_access/set_posix_access.
* security.h (set_file_attribute): Remove prototype.
(set_created_file_access): Add prototype.
(searchace): Ditto.
(set_posix_access): Ditto.
* syscalls.cc (open): Call open_with_arch with mode not umasked.
* sec_acl.cc: Change preceeding comment explaining new-style ACLs.
Describe how to generate deny ACEs in more detail. Accommodate the
fact that a NULL deny ACE is used for {DEF_}CLASS_OBJ, rather than
a special Cygwin ACE. Improve further comments.
(CYG_ACE_NEW_STYLE): Define.
(get_posix_access): Change from Cygwin ACE to NULL deny ACE. Fix
CLASS_OBJ handling to generate CLASS_OBJ and DEF_CLASS_OBJ from a single
NULL deny ACE if the inheritance flags say so.
* sec_helper.cc (well_known_cygwin_sid): Remove.
* security.h (well_known_cygwin_sid): Drop declaration.
* sec_acl.cc (CYG_ACE_ISBITS_TO_WIN): Fix typo.
(get_posix_access): Rename index variable from i to idx. Define only
once at top level.
* sec_acl.cc (get_posix_access): Handle multiple ACEs for the
owner and primary group of the file. Handle the default primary
group ACE as DEF_GROUP_OBJ entry if the directory has the S_ISGID bit
set. Add comments. Minor code rearrangements.
Preliminary read side implementation of new permission handling.
* acl.h (MAX_ACL_ENTRIES): Raise to 2730. Add comment to explain.
* sec_acl.cc: Add leading comment to explain new ACL style.
Add definitions and macros to use for bits in new Cygwin ACL.
(DENY_RWX): New mask value for all temporary deny bits.
(getace): Add bool parameter to decide when leaving all bits intact,
rather than filtering them per the already set bits.
(get_posix_access): New function, taking over functionality to read
POSIX ACL from SECURITY_DESCRIPTOR.
(getacl): Just call get_posix_access.
* sec_helper.cc (well_known_cygwin_sid): Define.
* security.cc (get_attribute_from_acl): Remove.
(get_info_from_sd): Remove.
(get_reg_sd): Call get_posix_access instead of get_info_from_sd.
(get_file_attribute): Ditto.
(get_object_attribute): Ditto.
* security.h (well_known_cygwin_sid): Declare.
(get_posix_access): Add prototype.
* Throughout, use simpler ACE macros from Windows' accctrl.h.
* getfacl.c (main): Special-case SYSTEM and Admins group. Add comments.
* setfacl.c: Align more to Linux tool.
(delacl): New function to delete acl entries only.
(modacl): Drop delete functionality. Add handling of recomputing the
mask and default mask values.
(delallacl): Rename from delacl.
(setfacl): Call delacl in Delete case. Call delallacl in DeleteAll
and DeleteDef case.
(usage): Accommodate new options. Rearrange and rephrase slightly.
(longopts): Emit 'x' in --delete case. Add --no-mask and --mask
options.
(opts): Add -x and -n options.
(main): Handle -d and -x the same. Handle -n and --mask options.
Drop handling for -r option.
* getfacl.c (usage): Align more closely to Linux version. Add new
options -c, -e, -E. Change formatting to accommodate longer options.
(longopts): Rename --noname to --numeric. Keep --noname for backward
compatibility. Add --omit-header, --all-effective and --no-effective
options.
(opts): Add -c, -e and -E option.
(main): Handle new -c, -e, and -E options.
Corinna Vinschen [Wed, 18 Nov 2015 18:12:37 +0000 (19:12 +0100)]
Don't generate Windows native symlinks targeting Cygwin-only targets
* path.cc (symlink_native): Don't generate native symlink targeting
Cygwin special file. Inform the caller.
(symlink_worker): Don;t bail out in WSYM_nativestrict case if the
target is a Cygwin special file.
Marcus Shawcroft [Tue, 10 Nov 2015 14:08:42 +0000 (14:08 +0000)]
[ARM] Adding strlen.S
The patch adds strlen.S to contain the complementary preprocessor
logic to strlen-stub.c intended to provide #inclusion of alternative
.S implementations.
Initially we just include the existing strlen-armv7.S implementation.
We rewrite _ISA_ARMV7 in both strlen.S and strlen-stub.c to use the
underlying existing underlying defintion from arm_asm.h in order to
avoide including that file, this is in effect the first step towards a
move to ACLE predefines only.
Tested by building newlib and comparing libc.a binaries before and
after for all permutations of:
This patch flattens the condition code selection used in strlen in an
attempt to make the guarding condition for each alternative
implementation clearer and to structure the logic in a manner that
makes it easier to maintain complementary logic between the
alternative 'C' and assembler implementations.
Tested by building newlib and comparing libc.a binaries before and
after for all permutations of:
Anton Kolesov [Fri, 23 Oct 2015 18:23:40 +0000 (21:23 +0300)]
Remove obsolete ARC system
ARC architecture specific files has been added ages ago in newlib/libc/sys,
but with invention of libgloss those files should be moved from newlib to
libgloss.
newlib/ChangeLog:
2015-11-12 Anton Kolesov <Anton.Kolesov@synopsys.com>
Regression testing newlib in conjunction with libgloss and
--enable-multilib can result in incompatible multilib versions of
newlib and libgloss being used during link.
This manifests on ARM target when newlib regression is run using a
GCC configured using --with-multilib-list=aprofile
With this configuration many of the multilib variants built are
mutually incompatible.
The issue is that the newlib dejagnu foo iterates each multilib
variant and correctly chooses the appropriate newlib variant but
always chooses the root/base libgloss variant.
The implementation of newlib/testsuite/lib/flags.exp contains the
following fragment:
set target_build_path "$objdir/$multibuildtop.."
The effect of this fragment is to explicitly select the root version
of libgloss, irrespective of the current multilib.
Digging around in VC it appears that the original implementation of
multlib magic came into the tree back in 2002 with:
In this initial version of multilib support, newlib was multilib
capable, but libgloss was not multilib capable, hence the necessity to
explicitly select the root libgloss version.
Subsequently flags.exp was modified to support out of tree testing:
This change is orthogonal to this issue, its effect is to exit early
in none multilib configurations.
Subsequently libgloss gained --enable-multilib support, the relevant
change is:
https://sourceware.org/ml/newlib/2006/msg00440.html
commit 00a4b31ad08aef361c5d74125ece410b4c285975
Author: Jeff Johnston <jjohnstn@redhat.com>
Date: Wed May 10 20:51:41 2006 +0000
This change enabled multilib support throughout libgloss, but ommitted
to adjust the flag.exp behaviour which anchors the libgloss multilib
selection to the base version.
The attached patch adjusts flags.exp to select the current multilib
variant of libgloss.
2015-11-06 Marcus Shawcroft <marcus.shawcroft@arm.com>
* testsuite/lib/flags.exp (libgloss_link_flags): Drop multilibtop
from target_build_path.
Jon Turney [Fri, 6 Nov 2015 14:40:23 +0000 (14:40 +0000)]
makedoc: Don't interpret license text as a command
Align makedoc's iscommand() with it's documentation, and don't allow commands to
contain a space. A command is a line containing only a sequence of capital
letters or '_', followed by optional spaces.
This prevents "IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
PARTICULAR PURPOSE" (the only line in the license text which consists of only
capitals and spaces, without any punctuation) from being interepreted as a
makedoc command, leading to:
"Can't find IMPLIED"
"warning, IMPLIED is not recognised"
being emitted by makedoc, (which is normally un-noticed because makedoc's stderr
is redirected to a .ref file)
2015-11-06 Jon Turney <jon.turney@dronecode.org.uk>
* doc/makedoc.c (iscommand): Only allow commands to have trailing
spaces, not space separated words.
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
Joseph Myers [Thu, 12 Nov 2015 12:09:24 +0000 (12:09 +0000)]
Avoid .eh_frame in ARM newlib.
ARM newlib has various strcmp implementations that use .cfi_*
directives to generate unwind information.
The effect of this is that the generated objects contain .eh_frame
sections. However, ARM uses its own unwind info format, not
.eh_frame, which is generated by ARM-specific directives, not .cfi_*.
The .eh_frame sections are useless, but also not removed by strip and
may be loaded into memory at runtime.
This patch fixes this by using .cfi_sections .debug_frame (as in
glibc) so that the directives generate .debug_frame instead.
.debug_frame is useful for the debugger, can be removed by strip, and
is not loaded into memory at runtime.
Corinna Vinschen [Thu, 12 Nov 2015 10:48:30 +0000 (11:48 +0100)]
Fix crash at process init on 64 bit XP/2003 under WOW64
* init.cc (munge_threadfunc): Don't call NtQueryInformationThread with
ThreadQuerySetWin32StartAddress info class on XP 64 and Server 2003 64.
It crashes.
The patch cleans up the auto configury mechanism used to select
different implementations of memchr for various architecture versions.
The approach here is to remove the selection of memchr within automake
and instead use complimentary logic in memchr-stub.c and memchr.S to
choose between the gerneric memchr.c implementation or one of the
architecture specific implementations.
This patch also changes the selection criteria inline with the
previous proposal here:
Regressed for armv7-a armv5 armv8-a, correct selection of memcpy
implementation by manual inspection of a test program built for these
three architectures.
Marcus Shawcroft [Fri, 30 Oct 2015 15:14:53 +0000 (15:14 +0000)]
Reorganize memcpy selection.
This patch cleans up the auto configury mechanism used to select
different implementations of memcpy for various architecture versions.
The approach here is to remove the selection of memcpy within automake
and instead use complimentary logic in memcpy-stub.c and memcpy.S to
choose between the generic memcpy.c implemenation or one of the
architecture specific memcpy*.S implemenations.
Regressed for armv7-a armv5 armv8-a, correct selection of memcpy
implementation by manual inspection of a test program built for these
three architectures.
This revised patch flips the remaining preprocessor logic in
memcpy-stub.c to use ACLE defines as requested in the previous review
and removes the now disused HAVE_ARMV7A and HAVE_ARMV8A configure.in
support.
Fix iterating over pending signals if a signal doesn't have to be cleared
* sigproc.cc (pending_signals::clear): Yet another fix to fix the fix.
Actually iterate over the list of pending signals even if there's a
signal which doesn't have to be cleared. Other than that, revert loop
to it's former self as a while loop.
Corinna Vinschen [Fri, 30 Oct 2015 19:13:26 +0000 (20:13 +0100)]
Workaround bug in LocaleNameToLCID on Windows 10
* nlsfuncs.cc (__get_lcid_from_locale): Handle LocaleNameToLCID
returning LOCALE_CUSTOM_UNSPECIFIED instead of failing in case of
an unsupported locale on Windows 10.
Corinna Vinschen [Fri, 30 Oct 2015 15:05:43 +0000 (16:05 +0100)]
Fix sigwait and pthread_kill return values in case of error
* signal.cc (sigwait): Fix return value to reflect errno in case of
error according to POSIX. Never return EINTR.
* thread.cc (pthread_kill): Return errno if sig_send failed.
Qian Hong [Thu, 29 Oct 2015 08:17:46 +0000 (09:17 +0100)]
Check for correct funtion entry address in munge_threadfunc
* init.cc (munge_threadfunc): Check that we're actually replacing
the correct original function address on the stack.
* ntdll.h (enum _THREADINFOCLASS): Add ThreadQuerySetWin32StartAddress.
Corinna Vinschen [Fri, 23 Oct 2015 12:30:40 +0000 (14:30 +0200)]
Clear pending signals targeting exiting thread
* cygtls.cc (_cygtls::remove): Call remove_pending_sigs.
* cygtls.h (_cygtls::remove_pending_sigs): Declare.
* sigproc.cc (pending_signals::clear): Define new method taking a
_cygtls pointer argument. Drop pending signals for that thread.
(_cygtls::remove_pending_sigs): Call pending_signals::clear for this
thread.
Corinna Vinschen [Thu, 22 Oct 2015 12:22:07 +0000 (14:22 +0200)]
Fix length returned from sys_cp_wcstombs in case nwc > # of wchars
* strfuncs.cc (sys_cp_wcstombs): Always return number of multibytes
without trailing NUL as the documentation implies. Throughout Cygwin,
fix usage to align to this pattern.
* fhandler_process.cc (format_process_winexename): Drop trailing NUL
and LF from output.