This reinstates with one change commit 74017d229d5e ("Cygwin: mmap:
use 64K pages for bookkeeping"), which was reverted in commit
fd57eea5617a. The change is that in mmap_record::match, we now align
the record length to the 4K Windows page boundary rather than the 64K
Windows allocation granularity. The reason for this is explained in
an extensive comment in the code.
Addresses: https://cygwin.com/pipermail/cygwin-patches/2025q1/013240.html
Fixes: 74017d229d5e ("Cygwin: mmap: use 64K pages for bookkeeping")
Signed-off-by: Ken Brown <kbrown@cornell.edu>
It seems that current _cygtls::handle_SIGCONT() code sometimes falls
into a deadlock due to frequent TLS lock/unlock operation in the
yield() loop. With this patch, the yield() in the wait loop is placed
outside the TLS lock to avoid frequent TLS lock/unlock.
Fixes: 9ae51bcc51a7 ("Cygwin: signal: Fix another deadlock between main and sig thread")
Reviewed-by: Corinna Vinschen <corinna@vinschen.de>
Signed-off-by: Takashi Yano <takashi.yano@nifty.ne.jp>
Checking EPERM only makes sense if the file exists, so
let the EEXIST check change places with the EPERM check.
Add a debug statement to the EPERM condition.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Some checks in path_conv are checking for various properties
to generate a boolean value, mostly to indicate different
combinations of on-disk files and devices.
Simplify these checks and, especially, document them inline.
Drop the isdevice() check in favor of a new isondisk() check.
Fixes: 4fc922b2c8a5 ("Cygwin: POSIX msg queues: Convert mqd_t to a descriptor")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This simplifies further checks for on-disk-devices in places with
special handling for such files.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Validate the fd returned by cygheap_getfd operating on given mqd.
A release note is provided for 3.5.6.
Reported-by: Christian Franke <Christian.Franke@t-online.de>
Addresses: https://cygwin.com/pipermail/cygwin/2025-January/257090.html
Signed-off-by: Mark Geisert <mark@maxrnd.com>
Fixes: 46f3b0ce85a9 (Cygwin: POSIX msg queues: move all mq_* functionality into fhandler_mqueue)
The commit a22a0ad7c4f0 was not entirely correct. Even with the patch,
some hangs still occur. This patch overrides the previous commit along
with the patch that makes cygwait() reentrant, to fix these hangs.
Addresses: https://cygwin.com/pipermail/cygwin/2024-December/256954.html
Addresses: https://cygwin.com/pipermail/cygwin/2024-December/256987.html
Fixes: d243e51ef1d3 ("Cygwin: signal: Fix deadlock between main thread and sig thread")
Fixes: a22a0ad7c4f0 ("Cygwin: signal: Do not handle signal when __SIGFLUSHFAST is sent")
Reported-by: Daisuke Fujimura <booleanlabel@gmail.com>
Reported-by: Jeremy Drake <cygwin@jdrake.com>
Reviewed-by: Corinna Vinschen <corinna@vinshcen.de>
Signed-off-by: Takashi Yano <takashi.yano@nifty.ne.jp>
To allow cygwait() to be called in the signal handler, a locally
created timer is used instead of _cygtls::locals.cw_timer if it is
in use.
Co-Authored-By: Corinna Vinschen <corinna@vinschen.de>
Signed-off-by: Takashi Yano <takashi.yano@nifty.ne.jp>
The call to TranslateNameW in s4uauth can fail and the code
leaves the function indicating an error, if so. It just
misses to set errno in this case, so add that.
Fixes: 0fb497165f85 ("Cygwin: seteuid: use Kerberos/MsV1_0 S4U authentication by default")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
If the macros REENTRANT_SYSCALLS_PROVIDED and MISSING_SYSCALL_NAMES are defined some
_reent_*-functions are replaced by the system-call and this leads to compile-warning
or a runtime-failure.
* newlib/libc/stdio/fopen.c _open_r is replaces by open(), declared in <fcntl.h>
* newlib/libc/stdio64/fopen64.c ditto
* newlib/libc/stdio/rename.c _rename_r is rename() itself; i.e, fix recursion
Use "Cygwin Setup program" and add a link to the homepage in some
cases.
Reported-by: David Dyck <david.dyck@gmail.com>
Reviewed-by: Jon Turney <jon.turney@dronecode.org.uk>
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
When Microsoft started to claim the "sshd" service name, we had
to rename our service to "cygsshd" but we never updated the docs.
Make sure the docs reflect the changed service name and fix a
few bumpy expression.
Reported-by: Mario Emmenlauer <mario@emmenlauer.de>
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
POSIX®.1-2024 now defines posix_spawn_file_actions_addchdir and
posix_spawn_file_actions_addfchdir. Add these interfaces to spawn.h,
guarded as POSIX 202405 symbols.
Cygwin-only: Export them as aliases of the *_np counterparts. Bump
API minor.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
FS_COMPR_FL is the Linux-compatible macro name. Redefine
FS_COMPRESSED_FL as an alias of FS_COMPR_FL for backward
compatibility.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This reverts commit 74017d229d5e46867c8a3ec01bf653e4392bf14c.
After this commit, gdb exits with a fatal error on startup.
Addresses: https://cygwin.com/pipermail/cygwin-patches/2025q1/013240.html
Fixes: 74017d229d5e ("Cygwin: mmap: use 64K pages for bookkeeping")
Signed-off-by: Ken Brown <kbrown@cornell.edu>
The code merging permissions relies on `pos' being set to the number
of current entries in the local aclent_t buffer. Commit 0e6d36766c83
("Cygwin: get_posix_access: move umask masking to the end") moved that
code to run earlier, but neglected to move setting `pos' correctly
as well.
Make sure to set `pos' inside the code block, as well as in the
final array size check, so `pos' is set correctly where it belongs.
Fixes: 0e6d36766c83 ("Cygwin: get_posix_access: move umask masking to the end")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
umask handling for new file gets overriden by subsequent merging of
permissions in Windows-generated ACLs. Fix this by performing
umask masking after all other ACL manipulations.
Fixes: a8716448cecc ("Simplify "Windows-standard-like" permissions")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Add a second loop to the code snippet merging permissions in old-style
or Windows-generated ACLs. This loop fixes up default ACL permissions
created from ACEs which are valid for the directory itself, as well as
getting inherited to child objects.
The FULL_ACE bit utilized for this is removed from the ACE at the
end of the function, together with the temporary DENY bits.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This reverts commit 971d2dffea7848270aa9dfb5c14dcd946c8971c0.
This patch was supposed to fix lots of FAILs in our own
testsuite/winsup.api/ltp/umask03.c test. The reason was
that umask masking in get_posix_access when generating new files
was overriden by later code in the same function, merging
permissions in old-style or Windows generated ACLs.
However, the solution to skip merging was not the right thing,
because this breaks handling of Windows-generated ACLs.
Rather, umask masking should be performed pretty late, certainly
after merging permissions. This will be done by a followup patch.
Fixes: 971d2dffea78 ("Cygwin: get_posix_access: do not merge permissions for just created files")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Setting the a_id in the default:user and default:group entries to
the actual uid and gid of the current owner/group doesn't make
sense. Change to ACL_UNDEFINED_ID.
Fixes: bc444e5aa4ca ("Reapply POSIX ACL changes.")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
The permissions of entries for SYSTEM and the Administrators group
are not added to the CLASS_OBJ entry, so they don't set the class
perms to rwx all the time.
This shouldn't be done for default perms, otherwise the resulting
permissions when generating new files might be surprisingly restricted
for Admins and SYSTEM.
Fixes: bc444e5aa4ca ("Reapply POSIX ACL changes.")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
When generating a default:group (Windows: CREATOR GROUP) ACL entry,
make sure to copy over user perms to the new default group entry.
Fixes: bc444e5aa4ca ("Reapply POSIX ACL changes.")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
After the commit d243e51ef1d3, zsh sometimes hangs at startup. This
occurs because SIGCHLD, which should trigger sigsuspend(), is handled
in cygwait() that is used to wait for a wakeup event in sig_send(),
even when __SIGFLUSHFAST is sent. Despite __SIGFLUSHFAST being
required to return before handling the signal, this does not happen.
With this patch, if the signal currently being sent is __SIGFLUSHFAST,
do not handle the received signal and keep it asserted after the
cygwait() for the wakeup event. Apply the same logic to the cygwait()
in the retrying loop for WriteFile() as well.
Addresses: https://cygwin.com/pipermail/cygwin/2024-December/256954.html
Fixes: d243e51ef1d3 ("Cygwin: signal: Fix deadlock between main thread and sig thread")
Reported-by: Daisuke Fujimura <booleanlabel@gmail.com>
Reviewed-by: Corinna Vinschen <corinna@vinschen.de>
Signed-off-by: Takashi Yano <takashi.yano@nifty.ne.jp>
It was convenient to use pages of size 4K (Windows page size) for
bookkeeping when we were using filler pages. But all references to
filler pages were removed in commit ceda26c9d35b ("Cygwin: mmap:
remove __PROT_FILLER and the associated methods"), so this is no
longer necessary. Switch to using pages of size 64K (Windows
allocation granularity) for everything.
Signed-off-by: Ken Brown <kbrown@cornell.edu>
unlink_nt() and rename2 () both check for the FILE_SUPPORTS_OPEN_BY_FILE_ID
flag to use POSIX delete/rename semantics. Both erroneously check the flag
against the file attributes using has_attributes().
Given that this flag is a filesystem flag, check using fs_flags() instead.
Fixes: fe2545e9faaf ("Cygwin: don't use unlink/rename POSIX semantics on certain NTFS")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Save the result of mmap_record::find_unused pages, and then pass that
result to the appropriate version of mmap_record::map_pages. Add a
new parameter of type off_t to the latter to make this possible, and
change its return type from off_t to bool. This saves map_pages from
having to call find_unused_pages again.
Signed-off-by: Ken Brown <kbrown@cornell.edu>
Slightly simplify the code and add comments to explain what the
function does. Add a version of this method with a new reference
parameter "contains" that is set to true if the chunk of this
mmap_record contains the given address range.
Signed-off-by: Ken Brown <kbrown@cornell.edu>
After commit a0933cd17d19, access(_, X_OK) returns 0 if the user
holds SE_BACKUP_PRIVILEGE, even if the file's ACL denies execution
to the user. This is triggered by trying to open the file with
FILE_OPEN_FOR_BACKUP_INTENT.
Fix check_file_access() so it checks for X_OK without specifying
the FILE_OPEN_FOR_BACKUP_INTENT flag if the file is not a directory.
Rearrange function slightly and add comments for easier comprehension.
Fixes: a0933cd17d19 ("Cygwin: access: Correction for samba/SMB share")
Reported-by: Bruno Haible <bruno@clisp.org>
Co-authored-by: Takashi Yano <takashi.yano@nifty.ne.jp>
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Previously mmap with MAP_FIXED would fail with EINVAL on an attempt to
map an address range contained in the chunk of an existing mapping.
With this commit, mmap will succeed, provided the mappings are
anonymous, the MAP_SHARED/MAP_PRIVATE flags agree, and MAP_NORESERVE
is not set for either mapping.
Addresses: https://cygwin.com/pipermail/cygwin/2024-December/256901.html
Signed-off-by: Ken Brown <kbrown@cornell.edu>
This commit fixes two problems. The first is that
mmap_is_attached_or_noreserve would sometimes call VirtualAlloc with
MEM_COMMIT on address ranges that were not known to have MEM_RESERVE
status. These calls could fail, causing SIGBUS to be raised
incorrectly. See
https://cygwin.com/pipermail/cygwin-developers/2024-December/012725.html
for details. Fix this by calling VirtualAlloc only on the part of the
address range that's contained in the current mmap_record.
The second problem is that the code would sometimes break out of the
main loop without knowing whether attached mappings still occur later
in the mmap_list. This could cause SIGBUS to not be raised when it
should be. Fix this by using "continue" rather than "break". For
efficiency, introduce a boolean variable "nocover" that's set to true
if we discover that the address range cannot be covered by noreserve
mmap regions.
Addresses:
https://cygwin.com/pipermail/cygwin-developers/2024-December/012725.html
Fixes: 70e476d27be8 ("* include/cygwin/version.h: Bump DLL version to
1.7.0")
Signed-off-by: Ken Brown <kbrown@cornell.edu>
Reportedly, the maximum pid of 65536 is much too small in bigger
environments. Raise the maximum PID to 4194304, whihc is the maximum
pid on Linux (PID_MAX_LIMIT defined in include/linux/threads.h).
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
In testing whether the requested area is contained in an existing
mapped region, an incorrect condition was used due to a
misinterpretation of the u_addr and u_len variables.
Addresses: https://cygwin.com/pipermail/cygwin/2024-December/256913.html
Fixes: c68de3a262fe5 ("* mmap.cc (class mmap_record):
Declare new map_pages method with address parameter.")
Signed-off-by: Ken Brown <kbrown@cornell.edu>
Previously, when unused pages from an mmap_record were recycled, they
were given the protection of the mmap_record rather than the
protection requested in the mmap call. Fix this by adding a
"new_prot" parameter to mmap_list::try_map() and
mmap_record::map_pages() to keep track of the requested protection.
Then use new_prot in the calls to VirtualProtect().
Addresses: https://cygwin.com/pipermail/cygwin/2024-December/256911.html
Fixes: f90e23f2714cb ("*autoload.cc (NtCreateSection): Define.")
Signed-off-by: Ken Brown <kbrown@cornell.edu>
If a realtime policy is selected, set the '(18) priority' field to the
negated sched_priority minus one. If SCHED_IDLE is selected, set it to
the lowest priority 39. Also set '(19) nice' to the originally requested
nice value. Ensure consistence with the current Windows priority in all
cases. Move the sched_priority from/to Windows priority mapping from
sched_get/setparam() to new functions in miscfuncs.cc.
Signed-off-by: Christian Franke <christian.franke@t-online.de>
In my build the below are treated as error now and causing failures. I
have described the fixes of each warning below.
In newlib/libc/sys/or1k/mlock.c:
CC libc/sys/or1k/libc_a-mlock.o
newlib/libc/sys/or1k/mlock.c: In function ‘__malloc_lock’:
newlib/libc/sys/or1k/mlock.c:56:19: warning: implicit declaration of function ‘or1k_critical_begin’ [-Wimplicit-function-declaration]
56 | restore = or1k_critical_begin();
| ^~~~~~~~~~~~~~~~~~~
newlib/libc/sys/or1k/mlock.c: In function ‘__malloc_unlock’:
newlib/libc/sys/or1k/mlock.c:93:17: warning: implicit declaration of function ‘or1k_critical_end’ [-Wimplicit-function-declaration]
93 | or1k_critical_end(restore);
| ^~~~~~~~~~~~~~~~~
This patch adds prototypes for functions or1k_critical_begin and
or1k_critical_end to suppress the warning, inline with what we do for
or1k_sync_cas.
In libgloss/or1k/or1k_uart.c:
libgloss/or1k/or1k_uart.c: In function ‘or1k_uart_set_read_cb’:
libgloss/or1k/or1k_uart.c:163:25: warning: passing argument 2 of ‘or1k_interrupt_handler_add’ from incompatible pointer type [-Wincompatible-pointer-types]
163 | _or1k_uart_interrupt_handler, 0);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
| |
| void (*)(uint32_t) {aka void (*)(long unsigned int)}
In file included from libgloss/or1k/or1k_uart.c:19:
libgloss/or1k/include/or1k-support.h:97:45: note: expected ‘or1k_interrupt_handler_fptr’ {aka ‘void (*)(void *)’} but argument is of type ‘void (*)(uint32_t)’ {aka ‘void (*)(long unsigned int)’}
97 | or1k_interrupt_handler_fptr handler,
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~
The public API is ‘void (*)(void *)' for our interrupt handlers. The
function _or1k_uart_interrupt_hander is the internal default
implementation of the uart IRQ handler and it doesn't use the data
argument.
This patch updates the _or1k_uart_interrupt_handler argument type from
uint32_t to void* allowing the function prototype to match the required
prototype.
If we did have a 64-bit implementation it would be an ABI issue. But,
there never has been one, or1k is only 32-bit.
In libgloss/or1k/interrupts.c:
libgloss/or1k/interrupts.c: In function ‘or1k_interrupt_handler_add’:
libgloss/or1k/interrupts.c:41:52: warning: assignment to ‘void *’ from ‘long unsigned int’ makes pointer from integer without a cast [-Wint-conversion]
41 | _or1k_interrupt_handler_data_ptr_table[id] = (uint32_t) data_ptr;
| ^
The table _or1k_interrupt_handler_data_ptr_table is an array of void*
and data_ptr is void*. There is no need for the cast so remove it.
In libgloss/or1k/sbrk.c:
libgloss/or1k/sbrk.c:23:29: warning: initialization of ‘uint32_t’ {aka ‘long unsigned int’} from ‘uint32_t *’ {aka ‘long unsigned int *’} makes integer from pointer without a cast [-Wint-conversion]
23 | uint32_t _or1k_heap_start = &end;
|
This patch adds a cast, which is safe in or1k as the architecture in
32-bit only. But this code would not be 64-compatible.
Signed-off-by: Stafford Horne <shorne@gmail.com>
