On Thu, Feb 13, 2014 at 06:43:15PM +0000, Ivaylo Dimitrov wrote: > > > On 13.02.2014 20:21, Will Deacon wrote: > > > > That's certainly unexpected. The n900 has (iirc) a Cortex-A8, which as an > > ARMv7 core, can perform unaligned accesses to normal, cacheable memory in > > hardware. > > > > Yep, Cortex-A8 and it has no problem to do unaligned memory accesses > AFAIK. I suspect it is a driver issue, not CPU. > > > Can you provide your .config and/or any information about your network chip > > please? There's a chance that the driver is doing something odd. > > The chip is TI wl1251, you can find the config file here(actually this > is the tree I am using to boot 3.14-rc1 on N900) - > https://gitorious.org/linux-n900/freemangordons-linux-n900/source/1434dbd7fbc5ec257b6cd6c547689b79177d1937:arch/arm/configs/rx51_defconfig
Ok, so based on that config I think we can narrow down the unaligned use to the following files: # Crypto crypto/memneq.c:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) crypto/memneq.c:#endif /* CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS */ crypto/memneq.c:#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS crypto/memneq.c:#endif /* CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS */ # Network include/linux/etherdevice.h:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) include/linux/etherdevice.h:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) include/linux/etherdevice.h:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) include/linux/etherdevice.h:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) include/linux/etherdevice.h:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) net/mac80211/rx.c:#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS net/ipv4/netfilter/arp_tables.c:#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS # Probably not relevant kernel/printk/printk.c:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) kernel/taskstats.c:#if defined(CONFIG_64BIT) && !defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) lib/lz4/lz4defs.h:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) \ lib/lz4/lz4defs.h:#else /* CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS */ lib/lzo/lzo1x_compress.c:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && defined(LZO_USE_CTZ64) lib/lzo/lzo1x_compress.c:#elif defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && defined(LZO_USE_CTZ32) lib/lzo/lzo1x_decompress_safe.c:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) lib/lzo/lzo1x_decompress_safe.c:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) lib/lzo/lzo1x_decompress_safe.c:#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) lib/strncpy_from_user.c:#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS lib/zlib_inflate/inffast.c:#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS Can you try hacking crypto/memneq.c so that it doesn't use CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS please? That would at least point the finger at net/mac80211/rx.c or similar. Cheers, Will -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
