Re: Extend faq.using to discuss fork failures

[Corinna Vinschen <corinna-cygwin at cygwin dot com>]

Hi Ryan,

Thanks for the FAQ entry.  I had a look now, finally.  Two nits:

On Aug 25 22:08, Ryan Johnson wrote:
> Index: winsup/doc/faq-using.xml
> ===================================================================
> RCS file: /cvs/src/src/winsup/doc/faq-using.xml,v
> retrieving revision 1.35
> diff -u -r1.35 faq-using.xml
> --- winsup/doc/faq-using.xml	4 Aug 2011 18:25:41 -0000	1.35
> +++ winsup/doc/faq-using.xml	26 Aug 2011 01:58:44 -0000
> @@ -1199,3 +1199,92 @@
>  </listitem>
>  </itemizedlist></para>
>  </answer></qandaentry>
> +<qandaentry id='faq.using.fixing-fork-failures'>
> +  <question><para>Calls to <literal>fork</literal> fail a lot. How can
> +  I fix the problem?</para></question>
> +  <answer>
> +
> +  <para>Unix-like applications make extensive use of
> +  <literal>fork</literal>, a function which spawns an exact copy of
> +  the running process. Notable fork-using applications include bash
> +  (and bash scripts), emacs, gcc, make, perl, python, and
> +  ruby. Unfortunately, the Windows ecosystem is quite hostile to a
> +  reliable fork implementation, leading to error messages such as:</para>
> +  <para><itemizedlist>
> +    <listitem>unable to remap <emphasis>$dll</emphasis> to same address as parent</listitem>
> +    <listitem>couldn't allocate heap </listitem>
> +    <listitem>died waiting for dll loading </listitem>
> +    <listitem>child -1 - died waiting for longjmp before initialization</listitem>
> +    <listitem>STATUS_ACCESS_VIOLATION </listitem>
> +    <listitem>resource temporarily unavailable </listitem>
> +  </itemizedlist></para>
> +  <para>If you find that frequent fork failures interfere with normal
> +  use of cygwin, please try the following: </para>
> +  <para><itemizedlist>
> +    <listitem>Restart whatever process is trying (and failing) to use
> +    <literal>fork</literal>. Sometimes Windows sets up a process
> +    environment that is even more hostile to fork than usual.</listitem>
> +    <listitem>Ensure that you have eliminated (not just disabled) all
> +    software on the BLODA (see <ulink
> +    url="http://cygwin.com/faq/faq.using.html#faq.using.bloda";
> +    />)</listitem>
> +    <listitem>Install the 'rebase' package, read its README in
> +    <literal>/usr/share/doc/Cygwin</literal>, and follow the
> +    instructions there to run 'rebaseall'.</listitem>

The rebase package is always installed since it's part of the Base
category.  This entry should be rephrased accordingly.

> +    </itemizedlist></para>
> +  <para>Please note that installing new packages or updating existing
> +  ones often undoes the effects of rebaseall and cause fork failures
> +  to reappear. If so, just run rebaseall again.
> +  </para></answer>
> +</qandaentry>
> +<qandaentry id='faq.using.why-fork-fails'>
> +  <question><para>Why does <literal>fork</literal> fail so much,
> +  anyway? (or: Why does <literal>fork</literal> still fail even though
> +  I ran rebaseall?)</para></question>
> +  <answer>
> +  <para>The semantics of <literal>fork</literal> require that a forked
> +  child process have <emphasis>exactly</emphasis> 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 <literal>fork</literal>
> +  implementation.

Everything else which follows from here is a good description of the
inner workings, but that shouldn't go into the FAQ.  What about creating
a link to the user's guide "Process Creation" section here.  The technical
details could then go into the "Process Creation" section.

> +  Three issues are especially prevalent:</para>
> +  <para><itemizedlist>
> +    <listitem>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 does 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
> +    resolved before <literal>cygwin1.dll</literal> 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 <literal>rebaseall</literal> package.</listitem>
> +
> +    <listitem>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 behavior
> +    interferes with a proper <literal>fork</literal>, 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). In a 64-bit system, marking
> +    executables as large address-ware and rebasing dlls to high
> +    addresses has been reported to help, as ASLR affects only the
> +    lower 2GB of address space.</listitem>
> +
> +    <listitem>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.</listitem></itemizedlist></para>
> +    <para>In summary, current Windows implementations make it
> +    impossible to implement a perfectly reliable fork, and occasional
> +    fork failures are inevitable. PTC.
> +    </para>
> +  </answer>
> +</qandaentry>


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat