First cut, still incomplete
* fhandler_socket is now base class for other socket classes
* fhandler_socket_inet handles AF_INET and AF_INET6 sockets
* fhandler_socket_local handles AF_LOCAL/AF_UNIX sockets
* finally get rid of fdsock by using set_socket_handle in accept4
* align file-related calls (fstat, fstatvfs, fchown, fchmod, facl)
to Linux.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
- Move definition of windows to POSIX error mapping struct into
cygerrno.h
- Move declaration of winsock errno functions to cygerrno.h
- Input to error mapping functions is DWORD
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Dropping the 'LL' specifier leads to 32 bit truncation during timestamp
computation. Revert it. Exempt MSPERSEC which is used for 32 bit values.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Add fhandler_socket::socket method
Add fhandler_socket::set_socket_handle method, basically duplicating
what fdsock is doing. This is the first step in getting rid of fdsock.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Windows does not remove FILE_ATTRIBUTE_TEMPORARY by itself after a
file has been closed. It's just some attribute which can be set or
removed at will, despite its purpose.
Apparently there are tools out there which use FILE_ATTRIBUTE_TEMPORARY
accidentally or wrongly, even Microsoft's own tools are affected. In
the end, the filesystem is potentially full of files with this attribute
set.
Implement O_TMPFILE files with FILE_ATTRIBUTE_TEMPORARY and
FILE_ATTRIBUTE_HIDDEN set. This combination is pretty unlikely.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* Fix an incorrect condition to recognize AF_LOCAL sockets in
file-related functions (fchmod, fchown, fstat, fsttavfs, facl, link).
* Return successfully when called on unnamed or abstract AF_LOCAL sockets,
except link, just as on Linux.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* Use 64 bit timestamps
* Use System boot and process start time to compute starttime value per
Linux proc.5 description.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
The previous patch introduced a compiler warning on x86.
Given time_t is only 4 bytes on x86 we get a long vs. unsigned long
comparison in timeval_to_ms. Fix it by careful casting.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* Redefine NSPERSEC to NS100PERSEC
* Define NSPERSEC as nanosecs per second
* Define USPERSEC as microsecs per second
* Use above constants throughout where appropriate
* Rename to_us to timespec_to_us and inline
* Rename it_bad to timespec_bad and inline
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Rearrange setsockopt/getsockopt into per level/per optname
preprocessing switch, actual call, per level/per optname
postprocessing switch for better readability as well as
extensibility.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Workaround the problem that bind doesn't fail with EADDRINUSE
if a socket with the same local address is still in TIME_WAIT.
Use IP Helper functions to check if such a socket exist and don't
even try this port, if so.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Assuming the address parameter is non-NULL, the test in
cygwin_bindresvport_sa only tests if the address family is
supported at all, which is insufficient.
Check if the incoming address family matches the socket
address family and for being AF_INET in cygwin_bindresvport
since the latter doesn't support any other family.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
We're still using ~SO_REUSEADDR because SO_EXCLUSIVEADDRUSE wasn't
defined in Mingw. But it is in Mingw-w64, so fix it.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This new function returns the name of the calling thread; works for both
cygthreads and pthreads. All calls to cygthread::name(/*void*/) replaced
by calls to mythreadname(/*void*/).
While POSIX mandates that certain socket types shall be defined by the
inclusing of <netinet/in.h>, it also says that this header may also make
visible all <sys/socket.h> symbols. Glibc does this, and without out it,
some packages end up requiring an additional #include <sys/socket.h>.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
POSIX does not mention the inclusion of <sys/time.h> in <sys/socket.h>
or <netinet/in.h>, nor is there anything in the latter two that would
require the former.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
When reading/writing block devices, Cygwin emulates Linux, providing
a byte-exact file position, albeit the underlying device drivers don't.
Unfortunately this only worked correctly for reading. The raw_write
method failed to revalidate the buffer after the read-modify-write
cycle in case len is not a multiple of the sector length. This in
turn resulted in lseek reporting a wrong file pointer.
Also, fix a condition for invalidating the buffer after writing from
a remaining read buffer.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Rounddown incoming addr on a page boundary. Without this, we may end
up with a fork error for private, anonymous maps. The reason is, we
use VirtualAlloc in this case which will potentially overcommit if
addr is not on a page boundary. This isn't taken into account in
bookkeeping, but fixup_mmaps_after_fork will eventually stumble over
this when trying to reproduce the copy-on-write pages: VirtualQuery
returns a region reaching beyond the supposedly allocated address
range and from there it goes downhill.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* change memcpy to internal _memcpy not setting the return value in %rax
* implement all memcpy-like functions as caller to _memcpy, setting %rax
to correct return value beforehand. This is possible because _memcpy
does not use %rax at all
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Since commit 8128f5482f, we have all the
non-tracing functions listed in posixoptions(7). The tracing functions
are gated by their own option, and are obsolecent anyway.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
* Don't use a bool var to store three states (-1, 0, 1).
* Correctly check for NT_SUCCESS of a function returning NTSTATUS.
* Straighten out code for better readability.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Commit 603ef545bd broke this snippet and
commit 5b312b4747 didn't help at all since
FILE_CREATE is exactly *not* the situation the test was originally
supposed to handle.
In fact, none of the open flags used by fhandler_base::open actually
hits this problem anymore, so just drop the code.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Difference to Linux: We can't create files which don't show up
in the filesystem due to OS restrictions. As a kludge, make a
(half-hearted) attempt to hide the file in the filesystem.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* Fix the maximum number of sockets allowed in the session to 2048,
instead of making it relative to sizeof(wsa_event).
The original choice of 2048 was in order to fit the wsa_events array
in the .cygwin_dll_common shared section, but there is still enough
room to grow there to have 2048 sockets on 64-bit as well.
* Return an error and set errno=ENOBUF if a socket can't be created
due to this limit being reached.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Also updates the fhandler_*::ftruncate implementations to adhere to the same
semantics. The error handling semantics of those syscalls that use
fhandler_*::ftruncate are moved to the implementations of those syscalls (
in particular ftruncate() and friends still set errno and return -1 on error
but that logic is handled in the syscall implementation).
Under some not quite clear conditions, NFS fails to use its
unlink workaround to rename a file to ".nfsXYZ". The problem has been
reproduced with the GAWK testext.awk testcase. To workaround this in
Cygwin, we now call try_to_bin on NFS, too. For some reason NFS doesn't
fail to rename the .cygXYZ file to .nfsXYZ after this Cygwin rename.
Fix comment in unlink_nt accordingly.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
The "final trick" code in try_to_bin accidentally never worked on
remote drives because it relies on rootdir. Which isn't set for
remote unlinks. The code now creates a full path for remote files.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
The try_to_bin function isn't called for netapp FSes anyway, so testing
for this FS type in the function is moot.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
The first argument of gethostbyaddr needs to accept a generic pointer
to be compatible with e.g. struct in_addr *. This caused an issue
compiling krb5-1.15.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
herror etc. are MISC, rcmd etc. are BSD, addrinfo functions are
POSIX.1-2001, except for IDN functionality which is GNU.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
When fork fails, we can use "%s" now with system_sprintf for the errmsg
rather than a (potentially too small) buffer for the format string.
* fork.cc (fork): Use "%s" with system_printf now.
With "%C" format string, argument may convert in up to MB_LEN_MAX bytes.
Relying on sys_wcstombs to add a trailing zero here requires us to
provide a large enough buffer.
* smallprint.c (__small_vsprintf): Use MB_LEN_MAX+1 bufsize for "%C".
The third argument of RtlLookupFunctionEntry actually is documented as
_Inout_opt_ for both x64 and ARM, although generic doc says _Out_ only.
* exceptions.cc (__unwind_single_frame): Initialize hist variable.
In order to avoid the year 2038 problem, define time_t to a signed
integer with at least 64-bits. The type for time_t can be forced to
long with the --enable-newlib-long-time_t configure option or with the
_USE_LONG_TIME_T system configuration define.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Always use the __-decorated form of an attribute name in public
headers, as the bareword form is in the user's namespace, and we
don't want compilation to break just because the user defines the
bareword to mean something else.
Signed-off-by: Eric Blake <eblake@redhat.com>
- pthread_mutex::lock now takes a PLARGE_INTEGER timeout pointer
and uses that in the call to cygwait.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
- Introduce inline helper pthread_convert_abstime. It converts
an absolute timespec to a Windows LARGE_INTEGER timestamp,
depending on the used clock.
- Use this function from pthread_cond_timedwait and semaphore::timedwait
- Merge semaphore::_wait and semaphore::_timedwait into single _wait
method, taking a LARGER_INTEGER timestamp.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This is copied from musl (MIT license). This is newer and more thorough
than that of FreeBSD currently shipped only on Cygwin.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
For historical reasons peek_console was calling the functions
PeekConsoleInputA and ReadConsoleInputA. However, these functions are
not working correctly under at least codepage 65001 (UTF-8) on systems
prior to Windows 10.
Use PeekConsoleInputW and ReadConsoleInputW instead, which work
correctly under all systems and all codepages.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
FillConsoleOutputCharacterA doesn't work correctly in codepage 65001
(UTF-8). Looks like the character conversion function from ascii char
to unicode char works incorrectly then. Use FillConsoleOutputCharacterW
instead.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
So far we had two functions checking the content of a reparse point,
readdir_check_reparse_point in fhandler_disk_file.cc for the sake of
readdir, and symlink_info::check_reparse_point for the sake of
generic path checking.
* Rename check_reparse_point_target helper to check_reparse_point_string
and convert to static function.
* Create new check_reparse_point_target helper containing the core
reparse point checking code
* Just call check_reparse_point_target from readdir_check_reparse_point
and symlink_info::check_reparse_point and only perform the unique
task in those functions.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
struct sigaction is POSIX.1-1990 but siginfo_t, which is used by its
sa_sigaction member, is POSIX.1b-1993. Therefore it needs to be guarded
as well, and as part of a union, the struct size is protected.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
When SA_RESTART is not set on a socket, a blocking send() that is
interrupted mid-transition by a signal should return success (and
report just how many bytes were actually transmitted).
The err variable used here was not always guaranteed to be set
correctly in the loop, so better to just remove it and call
WSAGetLastError() explicitly.
There are two common sigpause variants, both of which take an int argument.
If you request _XOPEN_SOURCE or _GNU_SOURCE, you get the System V version,
which removes the given signal from the process's signal mask; otherwise
you get the BSD version, which sets the process's signal mask to the given
value.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
So far Cygwin's readdir returned the inode number of a mount target
in d_ino, rather than the actual inode number of the mount point in
the underlying filesystem. This not only results in a performance
hit if the mount target is a remote FS, it is also not done on other
POSIX systems.
Remove the code evaluating the mount target inode number.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This patch fixes a minor compatibility issue w/ cygwin mount point handling in
readdir(), compared to equivalent behavior of Linux and MacOS. dentry.d_ino
should indicate the INO of the mount point itself, not the target volume root
folder.
Changed return type from readdir_check_reparse_point to uint8_t, to avoid
unnecessarily being implicitly cast to and from a signed int.
Renamed a related local variable "attr" to "oattr" that was eclipsing a member
variable with the same name.
Joe L.
Mingw-w64 (where the code has been taken from) has 4 byte longs
independently of the architecture but x86_64 Cygwin has 64 bit longs.
So use fistpll instead of fistpl on x86_64 Cygwin.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Workaround a bug (or undocumented behaviour) in LCMapStringW:
It's documented(*) that the cchDest parameter is a byte count with
LCMAP_SORTKEY, but a character count otherwise. But the docs don't
state what happens if you combine LCMAP_SORTKEY with LCMAP_BYTEREV.
Tests indicate that LCMAP_SORTKEY treats cchDest as byte count, but
then LCMAP_BYTEREV treats it as char count in the same call. So the
latter swaps twice as much bytes in the destination buffer than the
byte count it returns, which potentially results in writing past the
end of the given output buffer.
Solution: Don't specify LCMAP_BYTEREV in the LCMapStringW(LCMAP_SORTKEY)
call, rather byte swap afterwards.
(*) https://msdn.microsoft.com/en-us/library/windows/desktop/dd318702(v=vs.85).aspx
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This leads to excessive lag when stracing processes if the inferior
process checks the process table. The reason is that ppid isn't set
in the procinfo memory of the dynamically loading strace itself.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Add new SYMBOLIC_LINK_FLAG_ALLOW_UNPRIVILEGED_CREATE flag to
CreateSymbolicLinkW call when running on W10 1703 or later.
Don't do that on older versions to avoid ERROR_INVALID_PARAMETER.
Preliminary, needs testing. There's an off-chance that the
flag results in the same ERROR_INVALID_PARAMETER on 1703 if the
developer settings are not enabled.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
RtlGetNtVersionNumbers returns the build number with some upper bits
set for no apparent reason. The fact that RtlGetNtVersionNumbers is
undocumented doesn't exactly help.
Just filter out the upper WORD for now. If build numbers are in
danger to become 6 digit numbers, re-evaluate.
Always retrieve FileCompressionInformation for non-empty
files if FileStandardInformation returns 0 allocated blocks.
This fixes stat.st_blocks for files compressed with CompactOS method.
Signed-off-by: Christian Franke <franke@computer.org>
previous commit 4c90db7bc8 introduced
a compile time error because libm/common/s_infconst.c used the remove
__fmath, __dmath, and __ldmath union types.
Since this is very old, and unused for a very long time, just drop the
file and thus the __infinity constants entirely.
Exception: Cygwin exports __infinity from the beginning. There's a very,
VERY low probability that any existing executable or lib still uses this
constant, but we just keep it in for backward compat, nevertheless.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Explicitly format the contents of /proc/loadavg to avoid the decimal point
getting localized according to LC_NUMERIC. Using anything other than '.' is
wrong and breaks top.
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
v2:
autoload PerfDataHelper functions
Keep loadavg in shared memory
Guard loadavg access by a mutex
Initialize loadavg to the current load
v3:
Shared memory version bump isn't needed if we are only extending it
Remove unused autoload
Mark inititalized flags as NO_COPY for correct behaviour in fork child
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
The locale_t type is provided by <xlocale.h> on Linux, FreeBSD, and Darwin.
While, like on some of those systems, it is automatically included by
<locale.h> with the proper feature test macros, its presence under this
particular name is still presumed in real-world software.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
The original dll_init code was living under the wrong assumption that
dll_dllcrt0_1 and in turn dll_list::alloc will be called for each
LoadLibrary call. The same wrong assumption was made for
cygwin_detach_dll/dll_list::detach called via FreeLibrary.
In reality, dll_dllcrt0_1 gets only called once at first LoadLibrary
and cygwin_detach_dll once at last FreeLibrary.
In effect, reference counting for DLLs was completely broken after fork:
parent:
l1 = dlopen ("lib1"); // LoadLibrary, LoadCount = 1
l2 = dlopen ("lib1"); // LoadLibrary, LoadCount = 2
fork (); // LoadLibrary in the child, LoadCount = 1!
child:
dlclose (l1); // FreeLibrary actually frees the lib
x = dlsym (l2); // SEGV
* Move reference counting to dlopen/dlclose since only those functions
have to keep track of loading/unloading DLLs in the application context.
* Remove broken accounting code from dll_list::alloc and dll_list::detach.
* Fix error handling in dlclose.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Windows NUL device returns only the lower 32 bit of the number of
bytes written. Implement a fake write function to ignore the underlying
NUL device.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
So far we copy *_impure_ptr into _main_tls->local_clib if the child
process has been forked from a pthread. But that's not required.
The local_clib area of the new thread is on the stack and the stack
gets copied from the parent anyway (in frok::parent). So we only
have to make sure _main_tls is pointing to the right address and
do the simple post-fork thread init.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This test was broken from the start. It leads to creating a completely
new stack for the main thread of the child process when started from
the main thread of the parent. However, the main thread of a process
can easily running on a completely different stack, if the parent's main
thread was created by calling fork() from a pthread. For an example,
see https://cygwin.com/ml/cygwin/2017-03/msg00113.html
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
We use errno AKA _REENT->_errno since the last century and only set
_impure_ptr->_errno for backward compat. Stop that. Also, remove
the last check for _impure_ptr->_errno in Cygwin code.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Note that this always returns with dli_sname and dli_saddr set to NULL,
indicating no symbol matching addr could be found.
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
POSIX states as follows about pthread_cond_wait:
If a signal is delivered to a thread waiting for a condition variable,
upon return from the signal handler the thread resumes waiting for the
condition variable as if it was not interrupted, or it returns zero
due to spurious wakeup.
Cygwin so far employs the latter behaviour, while Linux and BSD employ
the former one.
Align Cygwin behaviour to Linux and BSD.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
There are certain, very obscure scenarios, which render the Windows
CWD handle inaccessible for reopening. An easy one is, the handle can
be NULL if the permissions of the CWD changed under the parent processes
feet.
Originally we just set errno and returned, but in case of init at
process startup that left the "posix" member NULL and subsequent
calls to getcwd failed with EFAULT.
We now check for a NULL handle and change the reopen approach
accordingly. If that doesn't work, try to duplicate the handle instead.
If duplicating fails, too, we set the dir handle to NULL and carry on.
This will at least set posix to some valid path and subsequent getcwd
calls won't fail. A NULL dir handle is ok, because we already do this
for virtual paths.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This patch alters the behaviour of dll_list::topsort to preserve the
order of dlopen'd units.
The load order of unrelated DLLs is reversed every time fork is called,
since dll_list::topsort finds the tail of the list and then unwinds to
reinsert items. My change takes advantage of what should be undefined
behaviour in dll_list::populate_deps (ndeps non-zero and ndeps and deps
not initialised) to allow the deps field to be initialised prior to the
call and appended to, rather than overwritten.
All DLLs which have been dlopen'd have their deps list initialised with
the list of all previously dlopen'd units. These extra dependencies mean
that the unwind preserves the order of dlopen'd units.
The motivation for this is the FlexDLL linker used in OCaml. The FlexDLL
linker allows a dlopen'd unit to refer to symbols in previously dlopen'd
units and it resolves these symbols in DllMain before anything else has
initialised (including the Cygwin DLL). This means that dependencies may
exist between dlopen'd units (which the OCaml runtime system
understands) but which Windows is unaware of. During fork, the
process-level table which FlexDLL uses to get the symbol table of each
DLL is copied over but because the load order of dlopen'd DLLs is
reversed, it is possible for FlexDLL to attempt to access memory in the
DLL before it has been loaded and hence it fails with an access
violation. Because the list is reversed on each call to fork, it means
that a subsequent call to fork puts the DLLs back into the correct
order, hence "even" invocations of fork work!
An interesting side-effect is that this only occurs if the DLLs load at
their preferred base address - if they have to be rebased, then FlexDLL
works because at the time that the dependent unit is loaded out of
order, there is still in memory the "dummy" DONT_RESOLVE_DLL_REFERENCES
version of the dependency which, as it happens, will contain the correct
symbol table in the data section. For my tests, this initially appeared
to be an x86-only problem, but that was only because the two DLLs on x64
should have been rebased.
Signed-off-by: David Allsopp <david.allsopp@metastack.com>
Match glibc behaviour to expose the public bswap_* macros only with an
explicity #include <byteswap.h>; #include'ing <endian.h> should not expose
them.
Signed-off-by: Yaakov Selkowitz <yselkowi@redhat.com>
Don't set SO_RCVBUF/SO_SNDBUF to fixed values, thus disabling autotuning.
Patch modeled after a patch suggestion from Daniel Havey <dhavey@gmail.com>
in https://cygwin.com/ml/cygwin-patches/2017-q1/msg00010.html:
At Windows we love what you are doing with Cygwin. However, we have
been getting reports from our hardware vendors that iperf is slow on
Windows. Iperf is of course compiled against the cygwin1.dll and we
believe we have traced the problem down to the function fdsock in
net.cc. SO_RCVBUF and SO_SNDBUF are being manually set. The comments
indicate that the idea was to increase the buffer size, but, this code
must have been written long ago because Windows has used autotuning
for a very long time now. Please do not manually set SO_RCVBUF or
SO_SNDBUF as this will limit your internet speed.
I am providing a patch, an STC and my cygcheck -svr output. Hope we
can fix this. Please let me know if I can help further.
Simple Test Case:
I have a script that pings 4 times and then iperfs for 10 seconds to
debit.k-net.fr
With patch
$ bash buffer_test.sh 178.250.209.22
usage: bash buffer_test.sh <iperf server name>
Pinging 178.250.209.22 with 32 bytes of data:
Reply from 178.250.209.22: bytes=32 time=167ms TTL=34
Reply from 178.250.209.22: bytes=32 time=173ms TTL=34
Reply from 178.250.209.22: bytes=32 time=173ms TTL=34
Reply from 178.250.209.22: bytes=32 time=169ms TTL=34
Ping statistics for 178.250.209.22:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 167ms, Maximum = 173ms, Average = 170ms
------------------------------------------------------------
Client connecting to 178.250.209.22, TCP port 5001
TCP window size: 64.0 KByte (default)
------------------------------------------------------------
[ 3] local 10.137.196.108 port 58512 connected with 178.250.209.22 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0- 1.0 sec 768 KBytes 6.29 Mbits/sec
[ 3] 1.0- 2.0 sec 9.25 MBytes 77.6 Mbits/sec
[ 3] 2.0- 3.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 3.0- 4.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 4.0- 5.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 5.0- 6.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 6.0- 7.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 7.0- 8.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 8.0- 9.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 9.0-10.0 sec 18.0 MBytes 151 Mbits/sec
[ 3] 0.0-10.0 sec 154 MBytes 129 Mbits/sec
Without patch:
dahavey@DMH-DESKTOP ~
$ bash buffer_test.sh 178.250.209.22
Pinging 178.250.209.22 with 32 bytes of data:
Reply from 178.250.209.22: bytes=32 time=168ms TTL=34
Reply from 178.250.209.22: bytes=32 time=167ms TTL=34
Reply from 178.250.209.22: bytes=32 time=170ms TTL=34
Reply from 178.250.209.22: bytes=32 time=169ms TTL=34
Ping statistics for 178.250.209.22:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 167ms, Maximum = 170ms, Average = 168ms
------------------------------------------------------------
Client connecting to 178.250.209.22, TCP port 5001
TCP window size: 208 KByte (default)
------------------------------------------------------------
[ 3] local 10.137.196.108 port 58443 connected with 178.250.209.22 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0- 1.0 sec 512 KBytes 4.19 Mbits/sec
[ 3] 1.0- 2.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 2.0- 3.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 3.0- 4.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 4.0- 5.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 5.0- 6.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 6.0- 7.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 7.0- 8.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 8.0- 9.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 9.0-10.0 sec 1.50 MBytes 12.6 Mbits/sec
[ 3] 0.0-10.1 sec 14.1 MBytes 11.7 Mbits/sec
The output shows that the RTT from my machine to the iperf server is
similar in both cases (about 170ms) however with the patch the
throughput averages 129 Mbps while without the patch the throughput
only averages 11.7 Mbps. If we calculate the maximum throughput using
Bandwidth = Queue/RTT we get (212992 * 8)/0.170 = 10.0231 Mbps. This
is just about what iperf is showing us without the patch since the
buffer size is set to 212992 I believe that the buffer size is
limiting the throughput. With the patch we have no buffer limitation
(autotuning) and can develop the full potential bandwidth on the link.
If you want to duplicate the STC you will have to find an iperf server
(I found an extreme case) that has a large enough RTT distance from
you and try a few times. I get varying results depending on Internet
traffic but without the patch never exceed the limit caused by the
buffering.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
The termios code doesn't handle erasing of multibyte characters
in canonical mode, it always erases a single byte. When entering
a multibyte character and then pressing VERASE, the input ends up
with an invalid character.
Following Linux we introduce the IUTF8 input flag now, set by
default. When this flag is set, VERASE or VWERASE will check
if the just erased input byte is a UTF-8 continuation byte. If
so, it erases another byte and checks again until the entire
UTF-8 character has been removed from the input buffer.
Note that this (just as on Linux) does NOT work with arbitrary
multibyte codesets. This only works with UTF-8.
For a discussion what happens, see
https://cygwin.com/ml/cygwin/2017-01/msg00299.html
Sidenote: The eat_readahead function is now member of fhandler_termios,
not fhandler_base. That's necessary to get access to the terminal's
termios flags.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
- Drop virtual_key_code (only used once)
- Convert macros wch and control_key_state to const vars
unicode_char and ctrl_key_state.
- Fix formatting
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Create two new inline functions is_alt_numpad_key(PINPUT_RECORD) and
is_alt_numpad_event(PINPUT_RECORD) which contain the actual checks.
Call these functions from fhandler_console::read and peek_console for
better readability.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
{p}select/{p}poll completely ignored Alt+Numpad key sequences in console
input which results in newer readline using pselect to fail handling such
sequences correctly. See https://cygwin.com/ml/cygwin/2017-01/msg00135.html
During debugging and testing it turned out that while reading console
input, single key presses during an Alt+Numpad sequences where not
ignored, so ultimately a sequence like
Alt-down Numpad-1 Numpad-2 Numpad-3
whihc is supposed to result in a single character in the input stream
will actually result in 4 chars in the input stream, three control
sequences and the actual character.
Both problems should be fixed by this patch.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
This patch fixes the following problem:
Commit 9636c426 refactored the pipe code especially to make sure
to call WriteFile only with chunks matching the maximum atomic write
count. This accidentally introduced a small change in behaviour
on blocking pipes due to the success case falling through into the
error case. Rather then writing atomic chunks until all bytes are
written, the code immediately broke from the loop after writing
the first chunk, basically the same as in case of non-blocking
writes. This behaviour is not compliant to POSIX which requires
"Write requests to a pipe or FIFO [...]
* If the O_NONBLOCK flag is clear, a write request may cause the
thread to block, but on normal completion it shall return nbyte."
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
If newfile already exists and is in use, trying to overwrite it with
NtSetInformationFile(FileRenameInformation) fails exactly as if we
don't have the permissions to delete it. Unfortunately the return code
is the same STATUS_ACCESS_DENIED, so we have no way to distinguish
these cases. What we do here so far is to start a transaction to delete
newfile. If this open fails with a transactional error we stop the
transaction and retry opening the file without transaction.
But, here's the problem: If newfile is in use, NtOpenFile(oldfile)
naturally does NOT fail with a transactional error. Rather, the
subsequent call to unlink_nt(newfile) does, because there's another
handle open to newfile outside a transaction. However, the code does
not check if unlink_nt fails with a transactional error and so fails
to retry without transaction.
This patch recifies the problem and checks unlink_nt's status as well.
Refactor code to get rid of goto into another code block.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
A *very* recent Windows build adds more code to the preamble of
RtlGetCurrentDirectory_U() so that the previous heuristic failed to find
the call to the locking routine.
This only affects the 64-bit version of ntdll, where the 0xe8 byte is
now found at offset 40, not the 32-bit version. However, let's just
double the area we search for said byte for good measure.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Corinna Vinschen
Mark Geisert
Yaakov Selkowitz
Xiaofeng Liu
Erik M. Bray
Yaakov Selkowitz
Michael Haubenwallner
Ken Brown
Sebastian Huber
Brian Inglis
Eric Blake
David Macek
Joe_Lowe
Jon Turney
Christian Franke
David Allsopp
Johannes Schindelin