[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: iceweasel on Fuloong (was: What is the status of Debian on Lemote Fuloong?)


On Mon, Jul 30, 2012 at 11:15:11PM +0200, Andreas Barth wrote:
> * Aaro Koskinen (aaro.koskinen@iki.fi) [120730 22:57]:
> > On Sat, Jul 28, 2012 at 10:48:46PM +0300, Aaro Koskinen wrote:
> > > madvise(0x6fc17000, 4096, MADV_DONTNEED) = -1 EINVAL (Invalid argument)
> > 
> > I believe this may be because the userspace assumes
> > wrong page size (4 KB). Kernel (since 2.6.32, commit
> > 315fe625f878749a7d2b6b65a40c29bbbe6e1dc7) supports only 16 KB and 64 KB
> > pages for this board.
> Do I understand it correct that this could be then an issue for any
> architecture, unless 4kb sizes are configured? (Which might be an
> non-common setup on other machines, but that's something else.)

Yes, it seems there is a Mozilla bug reported about this seen on PPC:


Comment #7 seems to provide a workaround that just skips the offending
code on PPC.

The issue I'm seeing is also caused by DecommitMemory:

#0  0x7792e940 in madvise () from /lib/mipsel-linux-gnu/libc.so.6
#1  0x772cc5a4 in js::gc::DecommitMemory (addr=<optimized out>, 
    size=<optimized out>)
    at /build/buildd-iceweasel_10.0.6esr-1-mipsel-22tkSa/iceweasel-10.0.6esr/js/src/jsgcchunk.cpp:370
#2  0x772c1854 in DecommitFreePages (cx=0x725990c0)
    at /build/buildd-iceweasel_10.0.6esr-1-mipsel-22tkSa/iceweasel-10.0.6esr/js/src/jsgc.cpp:2406
#3  SweepPhase (gckind=GC_SHRINK, gcmarker=0x7fa03cb0, cx=0x725990c0)
    at /build/buildd-iceweasel_10.0.6esr-1-mipsel-22tkSa/iceweasel-10.0.6esr/js/src/jsgc.cpp:2640
#4  MarkAndSweep (gckind=GC_SHRINK, cx=0x725990c0)
    at /build/buildd-iceweasel_10.0.6esr-1-mipsel-22tkSa/iceweasel-10.0.6esr/js/src/jsgc.cpp:2677
#5  GCCycle (cx=0x725990c0, comp=<optimized out>, gckind=GC_SHRINK)
    at /build/buildd-iceweasel_10.0.6esr-1-mipsel-22tkSa/iceweasel-10.0.6esr/js/src/jsgc.cpp:2921



Reply to: