Hi Jeff, Thanks for the patch. We have made several tests with a 2.6.18-1 kernel compiled with your patch and all the reported problems are gone.
Best regards and happy christmas, David > -----Mensaje original----- > De: Jeff Dike [mailto:[EMAIL PROTECTED] > Enviado el: martes, 27 de noviembre de 2007 20:13 > Para: Fermín Galán Márquez > CC: [EMAIL PROTECTED]; user-mode-linux-user@lists.sourceforge.net > Asunto: Re: [uml-user] Hanging using 'iptables -L' in UML > vm booted with con0=pts > > On Wed, Sep 12, 2007 at 11:52:32PM +0200, Fermín Galán > Márquez wrote: > > This is the trace (as you can see I're repeated the > experiment twice, > > the result is basically the same), hoping it could be useful to > > diagnose the problem and fix eventual bugs in the UML kernel. See > > below regarding the test procedure. > > > > (gdb) bt > > #0 0xb7f9d7f2 in ?? () from /lib/ld-linux.so.2 > > #1 0xb7f19ab0 in tcsetattr () from /lib/i686/cmov/libc.so.6 > > #2 0x080610d5 in generic_console_write (fd=7, > > buf=0x828fc13 "ip_tables: (C) 2000-2006 Netfilter > Core Team\n", n=45) > > at arch/um/drivers/chan_user.c:37 > > #3 0x0806096e in console_write_chan (chans=0x82756f0, > > buf=0x828fc13 "ip_tables: (C) 2000-2006 Netfilter > Core Team\n", len=45) > > at arch/um/drivers/chan_kern.c:352 > > My reaction to this has been that you are running UML under > a script which isn't reading UML output correctly, > therefore the pts device fills up, and UML hangs during > console output because the pts device buffer is full and > tcsetattr is waiting for it to empty. > > However, after looking at this, it turns up a host bug > where, under the conditions that UML is using it, tcsetattr > will always, guaranteed, return -EINTR. UML will then hang > because it will loop, making the call until it returns > something other than -EINTR. See > http://marc.info/?l=linux-kernel&m=119618990807182&w=2 for > my analysis. > > There is a workaround in UML, attached below. I basically > block SIGIO over the block of code which is fiddling > terminal attributes. > > Jeff > > -- > Work email - jdike at linux dot intel dot com > > Index: linux-2.6.22/arch/um/drivers/chan_user.c > =================================================================== > --- linux-2.6.22.orig/arch/um/drivers/chan_user.c > 2007-11-21 11:43:50.000000000 -0500 > +++ linux-2.6.22/arch/um/drivers/chan_user.c > 2007-11-27 14:03:10.000000000 -0500 > @@ -74,10 +74,16 @@ void generic_free(void *data) > > int generic_console_write(int fd, const char *buf, int n) { > + sigset_t old, no_sigio; > struct termios save, new; > int err; > > if (isatty(fd)) { > + sigemptyset(&no_sigio); > + sigaddset(&no_sigio, SIGIO); > + if (sigprocmask(SIG_BLOCK, &no_sigio, &old)) > + goto error; > + > CATCH_EINTR(err = tcgetattr(fd, &save)); > if (err) > goto error; > @@ -97,8 +103,11 @@ int generic_console_write(int fd, const > * Restore raw mode, in any case; we *must* ignore any > error apart > * EINTR, except for debug. > */ > - if (isatty(fd)) > + if (isatty(fd)) { > CATCH_EINTR(tcsetattr(fd, TCSAFLUSH, &save)); > + sigprocmask(SIG_SETMASK, &old, NULL); > + } > + > return err; > error: > return -errno; > ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ User-mode-linux-user mailing list User-mode-linux-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user
