On Tue, 27 Sep 2005, Atsushi Nemoto wrote: > >>>>> On Tue, 27 Sep 2005 10:21:17 -0400 (EDT), Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> said: > > stern> Yes I did. You can see it at > stern> https://lists.one-eyed-alien.net/pipermail/usb-storage/2005-September/001953.html > > Thank you. But 'kmalloc(US_SENSE_SIZE, GFP_KERNEL)' is not enough (at > least) for MIPS since some MIPS chips have 32 byte cacheline and > ARCH_KMALLOC_MINALIGN is 8 on linux-mips. > > Using 'max(dma_get_cache_alignment(), US_SENSE_SIZE)' would be OK. If that is so, it's a bug in linux-mips. ARCH_KMALLOC_MINALIGN is supposed to be at least as large as a cacheline. See this comment in mm/slab.c: /* * Enforce a minimum alignment for the kmalloc caches. * Usually, the kmalloc caches are cache_line_size() aligned, except when * DEBUG and FORCED_DEBUG are enabled, then they are BYTES_PER_WORD aligned. * Some archs want to perform DMA into kmalloc caches and need a guaranteed * alignment larger than BYTES_PER_WORD. ARCH_KMALLOC_MINALIGN allows that. * Note that this flag disables some debug features. */ and also this comment (referring to the kmalloc caches): /* * For performance, all the general caches are L1 aligned. * This should be particularly beneficial on SMP boxes, as it * eliminates "false sharing". * Note for systems short on memory removing the alignment will * allow tighter packing of the smaller caches. */ Alan Stern - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
