Date: Sat, 6 Jan 2007 14:42:48 +0100 From: Alexander Leidinger <Alexander@Leidinger.net> To: "Scot Hetzel" <swhetzel@gmail.com> Cc: freebsd-emulation@freebsd.org Subject: Re: linuxolator: amd64 Linux Test Project failures Message-ID: <20070106144248.4c566fc3@Magellan.Leidinger.net> In-Reply-To: <790a9fff0612311121l662736e0ud6d3220382af072a@mail.gmail.com> References: <790a9fff0612290911t5ae69715gd2bf0dda0f9228f2@mail.gmail.com> <20061229213509.GA86839@stud.fit.vutbr.cz> <790a9fff0612291424g4ecbd088i7846d248851b3e63@mail.gmail.com> <20061230120722.GA36814@stud.fit.vutbr.cz> <790a9fff0612301926n1562d560r1f3fff9af4bdb138@mail.gmail.com> <790a9fff0612301955r35b16472r2429fbe3fce78228@mail.gmail.com> <20061231144257.61520dc8@Magellan.Leidinger.net> <20061231142412.GA28462@stud.fit.vutbr.cz> <790a9fff0612311121l662736e0ud6d3220382af072a@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Quoting "Scot Hetzel" <swhetzel@gmail.com> (Sun, 31 Dec 2006 13:21:27 -0600): > On 12/31/06, Divacky Roman <xdivac02@stud.fit.vutbr.cz> wrote: > > On Sun, Dec 31, 2006 at 02:42:57PM +0100, Alexander Leidinger wrote: > > > Quoting "Scot Hetzel" <swhetzel@gmail.com> (Sat, 30 Dec 2006 21:55:37 -0600): > > > > > > > I think I found the problem, I had wrongly assumed that the code was > > > > using (char *)-1 as the address to pass to the function. Instead the > > > > code in access03, is using mmap to assign an address to variable > > > > bad_address. > > > > > > > > Below is the test program that shows that the problem is in mmap. > > > > > > http://lists.freebsd.org/pipermail/freebsd-emulation/2006-September/002625.html > > > > I guess what alexander wanted to say is that you should test the mmap patch > > that got commited to i386 mmap (see > > http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/i386/linux/linux_machdep.c.diff?r1=1.55&r2=1.56) > > > > and apply that to amd64 mmap and see if that fixes the issue you see.. > > > > pls do that and inform us > > > That patch to mmap fixes the problems with mmap on amd64, and brings > the failed LTP testcases closer to the i386 failed testcases. > > Results of mmap fingerprinter: > mmap1.txt - before mmap patch > mmap2.txt - after mmap patch I worry about this part: - /* - * XXX i386 Linux always emulator forces PROT_READ on (why?) - * so we do the same. We add PROT_EXEC to work around buggy - * applications (e.g. Java) that take advantage of the fact - * that execute permissions are not enforced by x86 CPUs. - */ - bsd_args.prot = linux_args->prot | PROT_EXEC | PROT_READ; + + bsd_args.prot = linux_args->prot; Can you please test if a java program (linux jvm, not a native jvm) like for example the linux version of eclipse or azureus is working on amd64? If it fails with the patch and without the patch, I will commit it. If it doesn't work after the patch but before the patch, we need to extend the mmap finger printing tool regarding the EXEC stuff. One more thought... if it fails with the patch but not without it: add the "|PROT_EXEC" part to the patched version. Bye, Alexander. -- The joys of love made her human and the agonies of love destroyed her. -- Spock, "Requiem for Methuselah", stardate 5842.8 http://www.Leidinger.net Alexander @ Leidinger.net: PGP ID = B0063FE7 http://www.FreeBSD.org netchild @ FreeBSD.org : PGP ID = 72077137
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070106144248.4c566fc3>