On Sun, Sep 29, 2013 at 10:02 PM, Markus Rechberger <mrechberger@xxxxxxxxx> wrote: > This patch adds memory mapping support to USBFS for isochronous and bulk > data transfers, it allows to pre-allocate usb transfer buffers. > > The CPU usage decreases 1-2% on my 1.3ghz U7300 notebook > The CPU usage decreases 6-8% on an Intel Atom n270 when > transferring 20mbyte/sec (isochronous), it should be more interesting to > see those > statistics on embedded systems where copying data is more expensive. X86 platform should be memory coherent, which means uncached mapping might not affect memory access performance from CPU, so you might need provide performance data on other non-coherent ARCH. > Usage from userspace: > allocation: > rv = ioctl(priv->usbfd, USBDEVFS_ALLOC_MEMORY, &mem); > if (rv == 0) > buffer = mmap(NULL, size, PROT_READ|PROT_WRITE, > MAP_SHARED, priv->usbfd, mem.offset); > use the mapped buffer with urb->buffer. > release: > rv = munmap(buffer, size); > memset(&mem, 0x0, sizeof(struct usbdevfs_memory)); > mem.buffer = buffer; > rv = ioctl(priv->usbfd, USBDEVFS_RELEASE_MEMORY, &mem); > > non freed memory buffers are collected and will be released when closing > the node. > > I tested this patch with Bulk and Isochronous, with and without memory > mapping (applications which don't use mmap will just fall back to the > legacy mechanism). > > Version 0.3: > * Removed comment > * Renamed USBDEVFS_FREE_MEMORY to USBDEVFS_RELEASE_MEMORY > * Clearing allocated memory > > Version 0.4: > * overriding SG transfers once memory mapped buffers are allocated for > BULK Why do you want to override SG transfer? > * adding pgprot_noncached to ensure that the IO memory is not cached Accessing non-cached mapping on some ARCHs is extremely slow, so I an wondering if it is good way to map the area as non-cached. Thanks, -- Ming Lei -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
