When a binary linked against the library is executed, the Cygwin DLL is loaded into the application's text segment. Because we are trying to emulate a UNIX kernel which needs access to all processes running under it, the first Cygwin DLL to run creates shared memory areas and global synchronization objects that other processes using separate instances of the DLL can access. This is used to keep track of open file descriptors and to assist fork and exec, among other purposes. Every process also has a per_process structure that contains information such as process id, user id, signal masks, and other similar process-specific information.
The DLL is implemented as a standard DLL in the Win32 subsystem. Under the hood it's using the Win32 API, as well as the native NT API, where appropriate.
Some restrictions apply for calls to the Win32 API. For details, see the section called “Restricted Win32 environment”, as well as the section called “Using the Win32 file API in Cygwin applications”.
The native NT API is used mainly for speed, as well as to access NT capabilities which are useful to implement certain POSIX features, but are hidden to the Win32 API.
Due to some restrictions in Windows, it's not always possible to strictly adhere to existing UNIX standards like POSIX.1. Fortunately these are mostly corner cases.
Note that many of the things that Cygwin does to provide POSIX compatibility do not mesh well with the native Windows API. If you mix POSIX calls with Windows calls in your program it is possible that you will see uneven results. In particular, Cygwin signals will not work with Windows functions which block and Windows functions which accept filenames may be confused by Cygwin's support for long filenames.
Windows NT includes a sophisticated security model based on Access
Control Lists (ACLs). Cygwin maps Win32 file ownership and permissions to
ACLs by default, on file systems supporting them (usually NTFS). Solaris
style ACLs and accompanying function calls are also supported.
The chmod call maps UNIX-style permissions back to the Win32 equivalents.
Because many programs expect to be able to find the
/etc/group files, we provide utilities
that can be used to construct them from the user and group information
provided by the operating system.
Users with Administrator rights are permitted to chown files. With version 1.1.3 Cygwin introduced a mechanism for setting real and effective UIDs. This is described in the section called “POSIX accounts, permission, and security”. As of version 1.5.13, the Cygwin developers are not aware of any feature in the Cygwin DLL that would allow users to gain privileges or to access objects to which they have no rights under Windows. However there is no guarantee that Cygwin is as secure as the Windows it runs on. Cygwin processes share some variables and are thus easier targets of denial of service type of attacks.
Cygwin supports both POSIX- and Win32-style paths, using either forward or back slashes as the directory delimiter. Paths coming into the DLL are translated from POSIX to native NT as needed. From the application perspective, the file system is a POSIX-compliant one. The implementation details are safely hidden in the Cygwin DLL. UNC pathnames (starting with two slashes) are supported for network paths.
The layout of this POSIX view of the Windows file system space is
stored in the
/etc/fstab file. Actually, there is a
/etc/fstab file as well as a user-specific
At startup the DLL has to find out where it can find the
/etc/fstab file. The mechanism used for this is simple.
First it retrieves it's own path, for instance
C:\Cygwin\bin\cygwin1.dll. From there it deduces
that the root path is
C:\Cygwin. So it looks for the
fstab file in
The layout of this file is very similar to the layout of the
fstab file on Linux. Just instead of block devices,
the mount points point to Win32 paths. An installation with
setup.exe installs a
fstab file by
default, which can easily be changed using the editor of your choice.
fstab file allows mounting arbitrary Win32
paths into the POSIX file system space. A special case is the so-called
It's the path under which every available drive in the system is mounted
under its drive letter. The default value is
so you can access the drives as
/cygdrive/d, etc... The cygdrive prefix can be set to
some other value (
/mnt for instance) in the
The library exports several Cygwin-specific functions that can be used by external programs to convert a path or path list from Win32 to POSIX or vice versa. Shell scripts and Makefiles cannot call these functions directly. Instead, they can do the same path translations by executing the cygpath utility program that we provide with Cygwin.
Win32 applications handle filenames in a case preserving, but case insensitive manner. Cygwin supports case sensitivity on file systems supporting that. Windows only supports case sensitivity when a specific registry value is changed. Therefore, case sensitivity is not usually the default.
Cygwin supports creating and reading symbolic links, even on Windows filesystems and OS versions which don't support them. See the section called “Symbolic links” for details.
Hard links are fully supported on NTFS and NFS file systems. On FAT and other file systems which don't support hardlinks, the call returns with an error, just like on other POSIX systems.
On file systems which don't support unique persistent file IDs (FAT,
older Samba shares) the inode number for a file is calculated by hashing its
full Win32 path. The inode number generated by the stat call always matches
the one returned in
d_ino of the
structure. It is worth noting that the number produced by this method is not
guaranteed to be unique. However, we have not found this to be a significant
problem because of the low probability of generating a duplicate inode number.
Cygwin supports Extended Attributes (EAs) via the linux-specific function
removexattr. All EAs
on Samba or NTFS are treated as user EAs, so, if the name of an EA is "foo"
from the Windows perspective, it's transformed into "user.foo" within Cygwin.
This allows Linux-compatible EA operations and keeps tools like
attr, or setfattr happy.
chroot is supported. Kind of. Chroot is not a
concept known by Windows. This implies some serious restrictions. First of
chroot call isn't a privileged call. Any user
may call it. Second, the chroot environment isn't safe against native windows
processes. Given that, chroot in Cygwin is only a hack which pretends security
where there is none. For that reason the usage of chroot is discouraged.
Don't use it unless you really, really know what you're doing.
It is often important that files created by native Windows applications be interoperable with Cygwin applications. For example, a file created by a native Windows text editor should be readable by a Cygwin application, and vice versa.
Unfortunately, UNIX and Win32 have different end-of-line conventions in text files. A UNIX text file will have a single newline character (LF) whereas a Win32 text file will instead use a two character sequence (CR+LF). Consequently, the two character sequence must be translated on the fly by Cygwin into a single character newline when reading in text mode.
This solution addresses the newline interoperability concern at the expense of violating the POSIX requirement that text and binary mode be identical. Consequently, processes that attempt to lseek through text files can no longer rely on the number of bytes read to be an accurate indicator of position within the file. For this reason, Cygwin allows you to choose the mode in which a file is read in several ways.
We chose to include Red Hat's own existing ANSI C library "newlib" as part of the library, rather than write all of the lib C and math calls from scratch. Newlib is a BSD-derived ANSI C library, previously only used by cross-compilers for embedded systems development. Other functions, which are not supported by newlib have been added to the Cygwin sources using BSD implementations as much as possible.
The reuse of existing free implementations of such things as the glob, regexp, and getopt libraries saved us considerable effort. In addition, Cygwin uses Doug Lea's free malloc implementation that successfully balances speed and compactness. The library accesses the malloc calls via an exported function pointer. This makes it possible for a Cygwin process to provide its own malloc if it so desires.
fork call in Cygwin is particularly interesting
because it does not map well on top of the Win32 API. This makes it very
difficult to implement correctly. Currently, the Cygwin fork is a
non-copy-on-write implementation similar to what was present in early
flavors of UNIX.
As the child process is created as new process, both the main
executable and all the dlls loaded either statically or dynamically have
to be identical as to when the parent process has started or loaded a dll.
While Windows does not allow to remove binaries in use from the file
system, they still can be renamed or moved into the recycle bin, as
outlined for unlink(2) in the section called “File Access related changes”.
To allow an existing process to fork, the original binary files need to be
available via their original file names, but they may reside in a
different directory when using the DotLocal (.local) Dll Redirection feature.
Since NTFS does support hardlinks, when the fork fails we try again, but
create a private directory containing hardlinks to the original files as
well as the
.local file now. The base directory for the
private hardlink directory is
you have to create manually for now if you need to protect fork against
updates to executables and dlls on your Cygwin instance. As hardlinks
cannot be used across multiple NTFS file systems, please make sure your
executable and dll replacing operations operate on the same single NTFS file
system as your Cygwin instance and the
directory. Note that this private hardlink directory also does help for
when a wrong dll is found in the parent process' current working directory.
To enable creating the hardlinks, you need to stop all currently running
Cygwin processes after creating this directory, once per Cygwin installation:
$ mkdir --mode=a=rwxt /var/run/cygfork
We create one hardlink directory per user, application and application age, and remove it when no more processes use that directory. To indicate whether a directory still is in use, we define a mutex name similar to the directory name. As mutexes are destroyed when no process holds a handle open any more, we can clean up even after power loss or similar: Both the parent and child process, at exit they lock the mutex with almost no timeout and close it, to get the closure promoted synchronously. If the lock succeeded before closing, directory cleanup is started: For each directory found, the corresponding mutex is created with lock. If that succeeds, the directory is removed, as it is unused now, and the corresponding mutex handle is closed.
Before fork, when about to create hardlinks for the first time, the mutex is opened and locked with infinite timeout, to wait for the cleanup that may run at the same time. Once locked, the mutex is unlocked immediately, but the mutex handle stays open until exit, and the hardlinks are created. It is fine for multiple processes to concurrently create the same hardlinks, as the result really should be identical. Once the mutex is open, we can create more hardlinks within this one directory without the need to lock the mutex again.
The first thing that happens when a parent process forks a child process is that the parent initializes a space in the Cygwin process table for the child. It then creates a suspended child process using the Win32 CreateProcess call. Next, the parent process calls setjmp to save its own context and sets a pointer to this in a Cygwin shared memory area (shared among all Cygwin tasks). It then fills in the child's .data and .bss sections by copying from its own address space into the suspended child's address space. After the child's address space is initialized, the child is run while the parent waits on a mutex. The child discovers it has been forked and longjumps using the saved jump buffer. The child then sets the mutex the parent is waiting on and blocks on another mutex. This is the signal for the parent to copy its stack and heap into the child, after which it releases the mutex the child is waiting on and returns from the fork call. Finally, the child wakes from blocking on the last mutex, recreates any memory-mapped areas passed to it via the shared area, and returns from fork itself.
While we have some ideas as to how to speed up our fork implementation by reducing the number of context switches between the parent and child process, fork will almost certainly always be inefficient under Win32. Fortunately, in most circumstances the spawn family of calls provided by Cygwin can be substituted for a fork/exec pair with only a little effort. These calls map cleanly on top of the Win32 API. As a result, they are much more efficient. Changing the compiler's driver program to call spawn instead of fork was a trivial change and increased compilation speeds by twenty to thirty percent in our tests.
However, spawn and exec present their own set of difficulties. Because there is no way to do an actual exec under Win32, Cygwin has to invent its own Process IDs (PIDs). As a result, when a process performs multiple exec calls, there will be multiple Windows PIDs associated with a single Cygwin PID. In some cases, stubs of each of these Win32 processes may linger, waiting for their exec'd Cygwin process to exit.
The semantics of
fork require that a forked
child process have exactly the same address
space layout as its parent. However, Windows provides no native
support for cloning address space between processes and several
features actively undermine a reliable
implementation. Three issues are especially prevalent:
DLL base address collisions. Unlike *nix shared
libraries, which use "position-independent code", Windows shared
libraries assume a fixed base address. Whenever the hard-wired
address ranges of two DLLs collide (which occurs quite often), the
Windows loader must "rebase" one of them to a different
address. However, it may not resolve collisions consistently, and
may rebase a different dll and/or move it to a different address
every time. Cygwin can usually compensate for this effect when it
involves libraries opened dynamically, but collisions among
statically-linked dlls (dependencies known at compile time) are
cygwin1.dll initializes and
cannot be fixed afterward. This problem can only be solved by
removing the base address conflicts which cause the problem,
usually using the
Address space layout randomization (ASLR). Starting with
Vista, Windows implements ASLR, which means that thread stacks,
heap, memory-mapped files, and statically-linked dlls are placed
at different (random) locations in each process. This behaviour
interferes with a proper
fork, and if an
unmovable object (process heap or system dll) ends up at the wrong
location, Cygwin can do nothing to compensate (though it will
retry a few times automatically).
DLL injection by BLODA. Badly-behaved applications which inject dlls into other processes often manage to clobber important sections of the child's address space, leading to base address collisions which rebasing cannot fix. The only way to resolve this problem is to remove (usually uninstall) the offending app.
In summary, current Windows implementations make it impossible to implement a perfectly reliable fork, and occasional fork failures are inevitable.
When a Cygwin process starts, the library starts a secondary thread for use in signal handling. This thread waits for Windows events used to pass signals to the process. When a process notices it has a signal, it scans its signal bitmask and handles the signal in the appropriate fashion.
Several complications in the implementation arise from the fact that the signal handler operates in the same address space as the executing program. The immediate consequence is that Cygwin system functions are interruptible unless special care is taken to avoid this. We go to some lengths to prevent the sig_send function that sends signals from being interrupted. In the case of a process sending a signal to another process, we place a mutex around sig_send such that sig_send will not be interrupted until it has completely finished sending the signal.
In the case of a process sending itself a signal, we use a separate semaphore/event pair instead of the mutex. sig_send starts by resetting the event and incrementing the semaphore that flags the signal handler to process the signal. After the signal is processed, the signal handler signals the event that it is done. This process keeps intraprocess signals synchronous, as required by POSIX.
Most standard UNIX signals are provided. Job control works as expected in shells that support it.
Socket-related calls in Cygwin basically call the functions by the
same name in Winsock, Microsoft's implementation of Berkeley sockets, but
with lots of tweaks. All sockets are non-blocking under the hood to allow
to interrupt blocking calls by POSIX signals. Additional bookkeeping is
necessary to implement correct socket sharing POSIX semantics and especially
for the select call. Some socket-related functions are not implemented at
all in Winsock, as, for example, socketpair. Starting with Windows Vista,
Microsoft removed the legacy calls
Recent versions of Cygwin now implement all these calls internally.
An especially troublesome feature of Winsock is that it must be initialized before the first socket function is called. As a result, Cygwin has to perform this initialization on the fly, as soon as the first socket-related function is called by the application. In order to support sockets across fork calls, child processes initialize Winsock if any inherited file descriptor is a socket.
AF_UNIX (AF_LOCAL) sockets are not available in Winsock. They are implemented in Cygwin by using local AF_INET sockets instead. This is completely transparent to the application. Cygwin's implementation also supports the getpeereid BSD extension. However, Cygwin does not yet support descriptor passing.
select function is another
call that does not map cleanly on top of the Win32 API. Much to our
dismay, we discovered that the Win32 select in Winsock only worked on
socket handles. Our implementation allows select to function normally
when given different types of file descriptors (sockets, pipes,
handles, and a custom /dev/windows Windows messages
Upon entry into the select function, the first operation is to sort the file descriptors into the different types. There are then two cases to consider. The simple case is when at least one file descriptor is a type that is always known to be ready (such as a disk file). In that case, select returns immediately as soon as it has polled each of the other types to see if they are ready. The more complex case involves waiting for socket or pipe file descriptors to be ready. This is accomplished by the main thread suspending itself, after starting one thread for each type of file descriptor present. Each thread polls the file descriptors of its respective type with the appropriate Win32 API call. As soon as a thread identifies a ready descriptor, that thread signals the main thread to wake up. This case is now the same as the first one since we know at least one descriptor is ready. So select returns, after polling all of the file descriptors one last time.