X-Spam-Check-By: sourceware.org Date: Thu, 23 Feb 2006 13:19:54 -0500 From: Christopher Faylor To: cygwin@cygwin.com Subject: Re: Hanging at GetModuleFileName in inside_kernel function Message-ID: <20060223181954.GG31464@trixie.casa.cgf.cx> Reply-To: cygwin@cygwin.com References: <4BFEBADC-51DD-40CD-9595-9337ABCBDD8E@rehley.net> <000301c638a4$dfc6fd70$a501a8c0@CAM.ARTIMI.COM> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <000301c638a4$dfc6fd70$a501a8c0@CAM.ARTIMI.COM> User-Agent: Mutt/1.5.11 Mailing-List: contact cygwin-help@cygwin.com; run by ezmlm Precedence: bulk List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: cygwin-owner@cygwin.com Mail-Followup-To: cygwin@cygwin.com Delivered-To: mailing list cygwin@cygwin.com On Thu, Feb 23, 2006 at 06:13:40PM -0000, Dave Korn wrote: >On 23 February 2006 16:20, Peter Rehley wrote: > >> Yeah, I saw that change, and I tried yesterdays snapshot but it still >> hung. I also did some more googling and found that someone submitted >> a patch a few years ago. The patch checked to see if it was inside >> the ntdll.dll by looking at the handle. >> http://www.cygwin.com/ml/cygwin-patches/2003-q2/msg00004.html >> >> I found this google too. >> http://blogs.msdn.com/oldnewthing/archive/2004/01/28/63880.aspx >> >> I'm going to try that patch today and see what happens. Christopher >> didn't apply it because it was a bandage and didn't really fix the >> bigger problem. > >That all fits right in with your diagnosis of the problem; a deadlock >of some kind I guess is the only thing that could make >GetModuleFileName hang. > >Have you got a stack backtrace showing the call chain when this problem >arises? You'd need to decode the stack backtrace on multiple threads, if this is really that kind of deadlock. Btw, FWIW, I am excruciatingly aware of this type of deadlock. I suffered from it greatly when I moved more of the fork code into the "DLLentry" part of cygwin to deal with problems caused by Corinna's streamliing of mmap. >We may be able to come up with a substitute for GMFN that would solve >the problem. Funny you should mention that. I was just thinking that I could use either use the technique that gdb uses to find shared library names or possibly use the stuff in hookapi.cc. Of course, *that* would be subject to races. So you'd potentially be trading races for deadlocks. cgf -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/