In dll_crt0_0, both threadinterface->Init and sigproc_init allocate
windows object handles using unpredictable memory regions, which may
collide with dynamically loaded dlls when they were relocated.
Ken Brown [Wed, 27 Mar 2019 21:29:21 +0000 (21:29 +0000)]
Cygwin: FIFO: implement clear_readahead
Make fhandler_base::clear_readahead virtual, and implement
fhandler_fifo::clear_readahead. This is called by
dtable::fixup_after_exec; it clears the readahead in each client.
Cygwin: fix: seteuid32() must return EPERM if privileges are not held.
Starting w/ the intro of S4U, seteuid32() calls lsaprivkeyauth(), then
s4uauth(). s4uauth calls LsaRegisterLogonProcess().
LsaRegisterLogonProcess fails w/ STATUS_PORT_CONNECTION_REFUSED, if the
proper privileges are not held.
Because of RtlNtStatusToDosError(), this status would be mapped to
ERROR_ACCESS_DENIED, which in turn would map to EACCES. Therefore it is
useless to add this status to errmap[] (errno.cc), as s4auauth() should
return EPERM as errno here (i.e. if process is not privileged).
Hence the kludge.
Before the intro of S4U, seteuid32() called lsaprivkeyauth(), then
lsaauth(), then create_token(). Before the intro of Vista, the latter
would have called NtCreateToken().
NtCreateToken() would have failed w/ STATUS_PRIVILEGE_NOT_HELD for a
process w/o the proper privileges. In that case, calling seteuid32()
would have returned EPERM (as required).
Since the intro of Vista, and if the process had been started from an
UNelevated shell, create_token() does NOT reach NtCreateToken()!
As create_token() failed to properly set errno in that case, calling
seteuid32() would return errno as set by lsaauth(), i.e. EACCES, not
in agreement w/ Posix (a bug which was present for years).
(lsaauth() called LsaRegisterLogonProcess() which would fail)
Ken Brown [Mon, 25 Mar 2019 23:06:10 +0000 (23:06 +0000)]
Cygwin: FIFO: add support for the duplex case
If a FIFO is opened with O_RDWR access, create the pipe with
read/write access, and make the first client have the handle of that
pipe as its I/O handle.
Adjust fhandler_fifo::raw_read to account for the result of trying to
read from that client if there's no data.
Ken Brown [Mon, 25 Mar 2019 23:06:09 +0000 (23:06 +0000)]
Cygwin: FIFO: avoid crashes when cloning a client
fhandler_fifo::clone called fhandler_base::clone on each client
fhandler. But those fhandlers are actually fhandler_fifo objects, so
when fhandler_base::clone calls copyto, it's actually
fhandler_fifo::copyto that gets called. This can lead to mysterious
crashes.
Fix this by simply calling clone (which translates to
fhandler_fifo::clone) on each client fhandler.
Ken Brown [Fri, 22 Mar 2019 19:30:41 +0000 (19:30 +0000)]
Cygwin: FIFO: update select
Add static functions peek_fifo, thread_fifo, start_thread_fifo, and
fifo_cleanup to select.cc. These are based on the corresponding pipe
functions, the main difference being that peek_fifo loops through the
connected clients to see if any of them have data available for
reading.
Add the fhandler_fifo methods select_read, select_write, and
select_except.
Add accessor methods get_nclients, get_handle, and is_connected that
are needed by peek_fifo.
Ken Brown [Fri, 22 Mar 2019 19:30:38 +0000 (19:30 +0000)]
Cygwin: FIFO: improve EOF detection
Add a hit_eof method that tries to detect whether any clients are
connected. Before concluding that there are none, it gives the
listen_client thread time to update the client data.
Ken Brown [Fri, 22 Mar 2019 19:30:37 +0000 (19:30 +0000)]
Cygwin: FIFO: allow multiple writers
Introduce a 'fifo_client_handler' structure that can be used by a
reader to communicate with a writer using an instance of the named
pipe. An fhandler_fifo opened for reading creates a thread that does
the following:
- maintains a list of fifo_client_handlers
- listens for_clients trying to connect
- creates new pipe instances as needed so that there's always at
least one available for connecting.
The pipe instances are initially created in blocking mode, but they
are set to be non-blocking after a connection is made.
fhandler_fifo::raw_read now loops through the connected clients and
reads from the first one that has data available.
New fhandler_fifo methods: add_client, listen_client,
listen_client_thread, check_listen_client_thread.
Replace the create_pipe method by create_pipe_instance, which allows
unlimited pipe instances.
New helper functions: create_event, set_pipe_non_blocking.
Ken Brown [Fri, 22 Mar 2019 19:30:36 +0000 (19:30 +0000)]
Cygwin: FIFO: stop using overlapped I/O
Make fhandler_fifo a derived class of fhandler_base instead of
fhandler_base_overlapped.
Replace the create_pipe macro, which is based on
fhandler_pipe::create, by new create_pipe and open_pipe methods.
These use NT functions instead of Win32 functions. Replace fifo_name
by get_pipe_name, which returns a pointer to a UNICODE_STRING.
Remove the fnevent macro, which would now be needed only once.
Add a raw_write method, adapted from fhandler_base::raw_write.
Corinna Vinschen [Wed, 27 Mar 2019 12:53:32 +0000 (13:53 +0100)]
Cygwin: winpids: Fix getting process multiple times
Switching to Cywin-only PIDs introduced a new problem when collecting
Cygwin processes for `ps -W': A process can show up multiple times
again, if the Cygwin procinfo has been opened for a just execing
process. The execed process then shows up twice, once as Cygwin
process, but with the wrong Windows PID of the execing process,
once as Windows-only process.
The mechanism used to exclude these stray processes didn't work with
the new Cygwin pid handling anymore. To fix this
* check if the incoming Windows PID is the same as the PID in the
procinfo. If not, we have the PID of the execing process while
procinfo was already changed,
* always check if the process has already been handled, not only
for processes we got a procinfo for,
* simplify adding pid to pidlist since pid is now always correct.
Andrew Stubbs [Mon, 25 Mar 2019 15:33:53 +0000 (15:33 +0000)]
Add missing includes.
These missing includes were causing build warnings, but also a real bug in
which the "size" parameter to "write" was being passed in 32-bit, whereas it
ought to be 64-bit. This led to intermittent bad behaviour.
Corinna Vinschen [Sat, 23 Mar 2019 16:50:00 +0000 (17:50 +0100)]
Cygwin: strace: print windows and cygwin pid in event output
strace only printed the Windows PID in event output so far.
Especially now that Windows and Cygwin PID are decoupled, the
strace user might like to see the Cygwin pid in event output as
well. However, at process startup, the process might not have
a Cygwin PID yet.
To mitigate this, always print the Windows PID and only add the
Cygwin pid if it exists.
Corinna Vinschen [Tue, 19 Mar 2019 19:56:11 +0000 (20:56 +0100)]
Cygwin: wcsxfrm_l: Only byte swap if dest size is > 0
commit c0d7d3e1a2fa96db15613cbd68a68c96966bc402 removed the usage of the
LCMAP_BYTEREV flag in the call to LCMapStringW to workaround a strange
bug in LCMapStringW. This patch didn't take a userspace call of
wcsxfrm{_l} with NULL buffer and 0 size to evaluate the required buffer
size into account. This introduced a crash trying to byte swap the
NULL buffer. This patch fixes that problem.
Andrew Stubbs [Mon, 18 Mar 2019 16:18:09 +0000 (16:18 +0000)]
AMD GCN: Implement circular buffering.
The GCN port outputs stdout and stderr via a shared-memory interface.
Previously the buffer was limited to 1000 write operations, which was enough
for testing purposes, but easy to exhaust.
This patch implements a new circular buffering system allowing a greater
amount of output. The interface must allow hundreds of hardware threads to
output simultaneously. The new limit is UINT32_MAX write operations.
Unfortunately, there's no way to tell if the host side has also been updated.
This code will misbehave unless the gcn-run from GCC is also updated (although
it's fine the other way around), but that patch has already been committed.
QEMU maintainers have asked for a specification of the nios2
semihosting interface. Since it's essentially a copy of the m68k
implementation, this patch adds a document for that target as well.
Corinna Vinschen [Tue, 12 Mar 2019 20:21:13 +0000 (21:21 +0100)]
Cygwin: ps: simplify code
Always use NtQuerySystemInformation(SystemProcessIdInformation).
This drops two code paths calling NtQueryInformationProcess or
GetModuleFileNameExW and only requires to open the process
to fetch system time info.
Corinna Vinschen [Tue, 12 Mar 2019 19:55:24 +0000 (20:55 +0100)]
Cygwin: ps: show *all* processes in ps -W output
There's a long-standing bug in ps -W. It only shows processes which ps
can open with PROCESS_QUERY_LIMITED_INFORMATION permissions. However,
that fails for a lot of system processes. Due to that, output is
basically restricted to processes in the same session, as well as Cygwin
processes... which isn't *quite* what ps -W was supposed to do.
Basically we only need to open the process to fetch the image name. If
that fails, utilize the undocumented SystemProcessIdInformation info
class introduced with Windows Vista, which allows to fetch the image
name by specifying the PID. Restructure the code a bit.
Corinna Vinschen [Tue, 12 Mar 2019 16:09:42 +0000 (17:09 +0100)]
Cygwin: loadavg: improve debugging of load_init
When logging in via ssh with an unprivileged account,
PdhAddEnglishCounter returns with status 0x800007D0,
PDH_CSTATUS_NO_MACHINE. We didn't find any workaround
but the changes to improve debugging output may help
in future. Using UNICODE instead of ANSI functions is
a result of trying to fix this problem.
Also drop the prototype workaround for PdhAddEnglishCounterA.
It's not required anymore since Mingw-w64's pdh.h catched up.
Corinna Vinschen [Tue, 12 Mar 2019 10:41:35 +0000 (11:41 +0100)]
Cygwin: fork/exec: Allow all users PROCESS_QUERY_LIMITED_INFORMATION
Create process with standard rights, plus
PROCESS_QUERY_LIMITED_INFORMATION for authenticated users. This
allows to fetch basic process information and thus /proc/<PID>/stat
to succeed on foreign processes.
While at it, fix formatting in CreateProcess calls.
Corinna Vinschen [Tue, 12 Mar 2019 10:34:50 +0000 (11:34 +0100)]
Cygwin: proc: let stat info always succeed
There's no good reason to return blank if some of the info
couldn't be collected. Drop useless call collecting
SystemProcessorPerformanceInformation. Always return some
valid start_time, even if we couldn't collect ProcessTimes.
Corinna Vinschen [Tue, 12 Mar 2019 10:17:11 +0000 (11:17 +0100)]
Cygwin: proc: return more useful cmdline
Creating /proc/<PID>/cmdline requires permissions to communicate
with the target process via its signal pipe. If that fails, the
output is "<defunct>" which doesn't make sense most of the time.
Rather, call format_process_exename in this case to get more useful
process name info, albeit not the full cmdline.
Corinna Vinschen [Mon, 11 Mar 2019 20:40:04 +0000 (21:40 +0100)]
Cygwin: fix permissions of winpid symlinks
The winpid symlinks got created with no query permissions, so
only admins could see all Cygwin processes. Create symlinks
so everyone has query permissions instead.
Canceling the timer thread runs under lock. The thread uses the same
lock to guard its timer_tracker struct access. If the timing is bad,
timer_settime or timer_delete grab the lock at the same time, the timer
expires. In the end, cancel waits for the thread sync while the thread
waits for ther lock to be released.
Fix this by not waiting for the thread sync under lock.
On setting the timer, the thread is accidentally only canceled when
disarming the timer. This leaks one thread per timer_settimer call.
Move the thread cancellation where it belongs.
Cygwin: seteuid: do not verify lsaprivkeyauth token
We don't support setting groups via /etc/groups anymore. Also, the
initgroups group list is created via S4U, so we have "Interactive" vs.
"Network" token, an artificial and entirely irrelevant difference.
So, "verifying" the lsaprivkeyauth token may lead to rejecting a prefectly
valid token. Just remove the verify_token call.
Cygwin: Revert attempting to unload user profile after use
Revert "Cywin: user profile: unload impersonation user profile on exit"
Revert "Cygwin: seteuid: allow inheriting impersonation user profile handle"
Revert "Cygwin: user profile: add debug output to unload_user_profile"
Revert "Cygwin: user profile: Make an effort to unload unused user profiles"
Cygwin: fork: fix child process permissions, take 3
Per MSDN VirtualQueryEx requires PROCESS_QUERY_INFORMATION.
Testing showed that PROCESS_QUERY_LIMITED_INFORMATION is sufficient
since Windows 8.1. The assumption that Windows 8 is the same as
Windows 8 was not correct, it requires PROCESS_QUERY_INFORMATION
as well.
Fix that by splitting the Windows 8 wincaps into one for Windows 8
and one for Windows 8.1. Set needs_query_information for Windows 8.
Cygwin: load_user_profile: Don't give primary domain to ldap
If the user domain is the primary domain, LDAP is supposed to
use the default naming context. This is accomplished by setting
domain name to NULL in the call to cyg_ldap::fetch_ad_account.
Cygwin: ldap: Fix overwriting domain when creating naming context
cyg_ldap::fetch_ad_account creates a naming context from the
incoming domain, if it's not NULL. The algorithm overwrites
dots with \0 in domain while creating the naming context, but
neglects to restore the dots.
Fix that by never overwriting the incoming domain name.
Corinna Vinschen [Thu, 28 Feb 2019 14:05:13 +0000 (15:05 +0100)]
Cygwin: fenv.h: Add feature test macros, fix values
- feenableexcept,fedisableexcept, fegetexcept are GNU-only
- fegetprec, fesetprec are Solaris, use __MISC_VISIBLE
- _feinitialise is Cygwin-internal only
- Replace self-named FP precision values to values from
http://www.open-std.org/jtc1/sc22//WG14/www/docs/n752.htm
as used by Solaris.
- Change return value of fesetprec to adhere to the above document
and Solaris.
- Document fegetprec, fesetprec as Solaris functions, not as GNU
functions
Cygwin: authentication: Always initialize domain info
...before calling any of its method. It's no safe bet that
it's already initialized when calling s4uauth and adding it
to load_user_profile certainly doesn't hurt.
In case of a local machine account login, pi.lpProfilePath points
to the buffer returned by NetUserGetInfo, but NetApiBufferFree
is called prior to calling LoadUserProfileW. Fix by copying over
usri3_profile to the local userpath buffer, just as in the AD case.
Cygwin: Disable creating case-sensitive folders by default
Inspecting the content of case-sensitive directories
on remote machines results in lots of errors like
disappearing diretories and files, file not found, etc.
Corinna Vinschen [Tue, 26 Feb 2019 09:46:05 +0000 (10:46 +0100)]
Cygwin: timerfd: add a sleep when being debugged
A sleep is required on Windows 10 64 bit only before calling
RegisterClassW in the timerfd thread, and only when running
under strace. One of the child processes inheriting the timerfd
descriptor will get a STATUS_FLOAT_INEXACT_RESULT exception inside
of msvcrt.dll. It's apparently some timing problem. It occurs
in 4 out of 5 runs under strace only. WOW64 and Windows 7 64 bit
don't have this problem.
Corinna Vinschen [Tue, 26 Feb 2019 09:19:08 +0000 (10:19 +0100)]
Cygwin: timerfd: reduce size of shared mem region to a single page
The share section was created using the PAGE_SIZE constant,
but PAGE_SIZE is 64K. Fix that by using wincap.page_size()
instead, which returns the desired actual page size of 4K.
Corinna Vinschen [Mon, 25 Feb 2019 19:58:12 +0000 (20:58 +0100)]
Cygwin: timerfd: rework implementation
timerfd_tracker and timerfd_shared classes:
- Just because handles are shared, we don't have to store them in
shared memory. Move share handles into timerfd_tracker class.
- Drop shared instance counter since it's not required anymore.
timerfd_shared only stores the actual timer data.
- Drop timerfd_shared::create, just set clock id.
- Drop timerfd_shared::dtor, it's not required anymore.
- Drop timerfd_tracker::close, just call dtor where required.
- Rename timerfd_tracker::increment_instances to timerfd_tracker::dup.
It's the only reason it exists...
- timerfd_tracker::dtor now checks the non-shared pointers for NULL
before attempting to close them.
- timerfd_tracker::dtor handles decrementing the local instance count
by itself.
- Add a method timerfd_tracker::init_fixup_after_fork_exec to set
non-shared pointers to NULL. Together with the dtor patches it
fixes a problem with close_on_exec timerfd descriptors.
- Fix a bug in handling the thread synchronization event. It's
actually nice to create it before using it...
- Drop using sec_none{_nih} in InitializeObjectAttributes. It's
an unnecessary roundabout route just to get a NULL pointer.
- Slightly rework timechange window handling.
- Add more comments to explain what happens.
fhandler_timerfd:
- Drop cnew macro, it just hides what happens.
- fhandler_timerfd::fixup_after_exec now calls
timerfd_tracker::init_fixup_after_fork_exec first, so a subsequent
call to timerfd_tracker::dtor only works on valid handles.
Corinna Vinschen [Sun, 24 Feb 2019 19:15:36 +0000 (20:15 +0100)]
Cygwin: timerfd: fix shared memory allocation in fork/exec
timerfd_tracker::fixup_after_fork_exec always tries to restore
the shared timer region at the same address as in the parent.
This is entirely unnecessary and wasn't intended, rather some
kind of copy/paste thinko. Fix that. Print NtMapViewOfSection
status code in api_fatal on failure for debugging.
Corinna Vinschen [Sat, 23 Feb 2019 16:22:44 +0000 (17:22 +0100)]
Cygwin: user profile: fetch roaming profile path via LDAP
Commit 649911fb40e45bc9a1ad8a3c28d90eec78c9cb7f avoids the
calls to NetUserGetGroups and NetUserGetLocalGroups since
these can take a lot of time. The same problem potentially
occurs when loading the user profile. The code fetches
the roaming profile path calling NetUserGetInfo, which also
can be rather slow.
To avoid this problem, fetch the profile patch using LDAP.
Also, don't bail out early if the user's registry hive already
exists. This may result in outdated information.
Corinna Vinschen [Fri, 22 Feb 2019 17:27:38 +0000 (18:27 +0100)]
Cygwin: s4uauth: allow to be called for identification only
s4uath was only callable to create an impersonation token so
far. Rework the function to allow creating an identification
token for informational purposes even from untrusted processes.
Take domainname and username instead of a passwd pointer to be
more multi-purpose.
Corinna Vinschen [Fri, 22 Feb 2019 10:04:34 +0000 (11:04 +0100)]
Cygwin: passwd/group: drop fetching case-correct group names from LDAP
Commit 4e34a39b5cdf4c3f889486b7460bea063e579d10 made sure all user and
group names are case-correct, but it introduced a hefty performance hit
on starting the first Cygwin process.
Adding an ldap call for each AD group in a user token takes its toll in
bigger AD environments with lots of groups in a user token. Real-life
example: 300 groups w/ roundtrip time to the LDAP server of 0.25 secs
per call...
Corinna Vinschen [Mon, 18 Feb 2019 20:45:34 +0000 (21:45 +0100)]
Cygwin: s4uauth: convert token to primary token
Up to Vista CreateProcessAsUser only worked with primary tokens,
so convert S4U impersonation token to primary token. MSDN still
documents it that way, but actually an impersonation token is
sufficient since Windows 7.
Corinna Vinschen [Mon, 18 Feb 2019 20:00:59 +0000 (21:00 +0100)]
Cygwin: re-enable create_token for older systems
Under WOW64 on 64 bit Windows 7, MsV1_0S4ULogon appears to be
unimplemented, probably under Vista as well. Re-enable
create_token method, to allow basic seteuid on W7 WOW64 and
Vista as well.
Corinna Vinschen [Mon, 18 Feb 2019 16:59:56 +0000 (17:59 +0100)]
Cygwin: s4uauth: make sure to fetch correct package id
for domain accounts we try KerbS4ULogon first, MsV1_0S4ULogon
second. But we only fetch the package id for the supporting
authentication package (Kerberos/MsV1_0) once at the start.
Duplicate LsaLookupAuthenticationPackage call and move into the
Kerb/MsV1_0 branches so that it fetches the correct package id
for the method we call next.
Curious enough this worked before. Apparently both methods
work with the MICROSOFT_KERBEROS_NAME_A package id. However,
requesting and using the right authentication package id is
the prudent thing to do.
This potentially allows to circumvent OpenSSHs user/group name matching,
unless the Admin knows to add every local user twice or to use patterns,
e.g.:
Corinna Vinschen [Mon, 18 Feb 2019 09:12:07 +0000 (10:12 +0100)]
Cygwin: mount: define binary mount as default
Commit c1023ee353705671aa9a8e4e1179022277add2aa changed the way
path_conv::binmode() works. Rather than returning three states,
O_BINARY, O_TEXT, 0, it only returned 2 states, O_BINARY, O_TEXT. Since
mounts are only binary if they are explicitely mounted binary by setting
the MOUNT_BINARY flag, textmode is default.
This introduced a new bug. When inheriting stdio HANDLEs from native
Windows processes, the fhandler and its path_conv are created from a
device struct only. None of the path or mount flags get set this way.
So the mount flags are 0 and path_conv::binmode() returned 0.
After the path_conv::binmode() change it returned O_TEXT since, as
explained above, the default mount mode is textmode.
Rather than just enforcing binary mode for path_conv's created from
device structs, this patch changes the default mount mode to binary:
Replace MOUNT_BINARY flag with MOUNT_TEXT flag with opposite meaning.
Drop all explicit setting of MOUNT_BINARY. Drop local set_flags
function, it doesn't add any value.
When recognizing a negative pid, optind is off by one. The
code correcting this has been erroneously removed by commit 8de660271fe75a6993f1c9888d24b824bb7f999d. Revert that.