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.
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
* adding pgprot_noncached to ensure that the IO memory is not cached
(thanks to Ming Lei for mentioning that)
http://sundtek.de/support/devio_mmap_v0.4.diff
Signed-off-by: Markus Rechberger <patches@xxxxxxxxxxx>
diff --git a/drivers/usb/core/devio.c b/drivers/usb/core/devio.c
index 737e3c1..a32fb70 100644
--- a/drivers/usb/core/devio.c
+++ b/drivers/usb/core/devio.c
@@ -69,6 +69,7 @@ struct dev_state {
spinlock_t lock; /* protects the async urb lists */
struct list_head async_pending;
struct list_head async_completed;
+ struct list_head memory_list;
wait_queue_head_t wait; /* wake up if a request completed */
unsigned int discsignr;
struct pid *disc_pid;
@@ -96,6 +97,16 @@ struct async {
u8 bulk_status;
};
+struct usb_memory {
+ struct list_head memlist;
+ int vma_use_count;
+ int usb_use_count;
+ u32 offset;
+ u32 size;
+ void *mem;
+ unsigned long vm_start;
+};
+
static bool usbfs_snoop;
module_param(usbfs_snoop, bool, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(usbfs_snoop, "true to log all usbfs traffic");
@@ -288,6 +299,9 @@ static struct async *alloc_async(unsigned int numisoframes)
static void free_async(struct async *as)
{
+ struct usb_memory *usbm = NULL, *usbm_iter;
+ unsigned long flags;
+ struct dev_state *ps = as->ps;
int i;
put_pid(as->pid);
@@ -297,8 +311,22 @@ static void free_async(struct async *as)
if (sg_page(&as->urb->sg[i]))
kfree(sg_virt(&as->urb->sg[i]));
}
+
+ spin_lock_irqsave(&ps->lock, flags);
+ list_for_each_entry(usbm_iter, &ps->memory_list, memlist) {
+ if (usbm_iter->mem == as->urb->transfer_buffer) {
+ usbm = usbm_iter;
+ break;
+ }
+ }
+ spin_unlock_irqrestore(&ps->lock, flags);
+
kfree(as->urb->sg);
- kfree(as->urb->transfer_buffer);
+ if (usbm == NULL)
+ kfree(as->urb->transfer_buffer);
+ else
+ usbm->usb_use_count--;
+
kfree(as->urb->setup_packet);
usb_free_urb(as->urb);
usbfs_decrease_memory_usage(as->mem_usage);
@@ -811,6 +839,7 @@ static int usbdev_open(struct inode *inode, struct file *file)
INIT_LIST_HEAD(&ps->list);
INIT_LIST_HEAD(&ps->async_pending);
INIT_LIST_HEAD(&ps->async_completed);
+ INIT_LIST_HEAD(&ps->memory_list);
init_waitqueue_head(&ps->wait);
ps->discsignr = 0;
ps->disc_pid = get_pid(task_pid(current));
@@ -839,6 +868,8 @@ static int usbdev_release(struct inode *inode, struct file *file)
struct dev_state *ps = file->private_data;
struct usb_device *dev = ps->dev;
unsigned int ifnum;
+ struct list_head *p, *q;
+ struct usb_memory *tmp;
struct async *as;
usb_lock_device(dev);
@@ -863,6 +894,14 @@ static int usbdev_release(struct inode *inode, struct file *file)
free_async(as);
as = async_getcompleted(ps);
}
+
+ list_for_each_safe(p, q, &ps->memory_list) {
+ tmp = list_entry(p, struct usb_memory, memlist);
+ list_del(p);
+ if (tmp->mem)
+ free_pages_exact(tmp->mem, tmp->size);
+ kfree(tmp);
+ }
kfree(ps);
return 0;
}
@@ -1173,6 +1212,7 @@ static int proc_do_submiturb(struct dev_state *ps, struct usbdevfs_urb *uurb,
struct usb_host_endpoint *ep;
struct async *as = NULL;
struct usb_ctrlrequest *dr = NULL;
+ struct usb_memory *usbm = NULL, *iter = NULL;
unsigned int u, totlen, isofrmlen;
int i, ret, is_in, num_sgs = 0, ifnum = -1;
void *buf;
@@ -1259,9 +1299,13 @@ static int proc_do_submiturb(struct dev_state *ps, struct usbdevfs_urb *uurb,
goto interrupt_urb;
}
uurb->number_of_packets = 0;
- num_sgs = DIV_ROUND_UP(uurb->buffer_length, USB_SG_SIZE);
- if (num_sgs == 1 || num_sgs > ps->dev->bus->sg_tablesize)
- num_sgs = 0;
+
+ /* do not use SG buffers when memory mapped segments are allocated */
+ if (list_empty(&ps->memory_list)) {
+ num_sgs = DIV_ROUND_UP(uurb->buffer_length, USB_SG_SIZE);
+ if (num_sgs == 1 || num_sgs > ps->dev->bus->sg_tablesize)
+ num_sgs = 0;
+ }
break;
case USBDEVFS_URB_TYPE_INTERRUPT:
@@ -1323,6 +1367,8 @@ static int proc_do_submiturb(struct dev_state *ps, struct usbdevfs_urb *uurb,
goto error;
}
+ as->ps = ps;
+
u += sizeof(struct async) + sizeof(struct urb) + uurb->buffer_length +
num_sgs * sizeof(struct scatterlist);
ret = usbfs_increase_memory_usage(u);
@@ -1360,21 +1406,47 @@ static int proc_do_submiturb(struct dev_state *ps, struct usbdevfs_urb *uurb,
totlen -= u;
}
} else if (uurb->buffer_length > 0) {
- as->urb->transfer_buffer = kmalloc(uurb->buffer_length,
- GFP_KERNEL);
- if (!as->urb->transfer_buffer) {
- ret = -ENOMEM;
- goto error;
+ if (!list_empty(&ps->memory_list)) {
+ unsigned long flags;
+ usbm=NULL;
+ as->urb->transfer_buffer = NULL;
+ spin_lock_irqsave(&ps->lock, flags);
+ list_for_each_entry(iter, &ps->memory_list, memlist) {
+ if (iter->vm_start == (unsigned long)uurb->buffer && iter->usb_use_count == 0 &&
+ (PAGE_SIZE<<get_order(iter->size)) >= uurb->buffer_length) {
+ usbm = iter;
+ usbm->usb_use_count++;
+ break;
+ }
+ }
+ spin_unlock_irqrestore(&ps->lock, flags);
+ if (usbm) {
+ as->urb->transfer_buffer = usbm->mem;
+ } else {
+ ret = -ENOMEM;
+ goto error;
+ }
+ if (as->urb->transfer_buffer == NULL) {
+ ret = -ENOMEM;
+ goto error;
+ }
+ } else {
+ as->urb->transfer_buffer = kmalloc(uurb->buffer_length,
+ GFP_KERNEL);
+ if (!as->urb->transfer_buffer) {
+ ret = -ENOMEM;
+ goto error;
+ }
}
- if (!is_in) {
+ if (!is_in && usbm == NULL) {
if (copy_from_user(as->urb->transfer_buffer,
uurb->buffer,
uurb->buffer_length)) {
ret = -EFAULT;
goto error;
}
- } else if (uurb->type == USBDEVFS_URB_TYPE_ISO) {
+ } else if (uurb->type == USBDEVFS_URB_TYPE_ISO && usbm == NULL) {
/*
* Isochronous input data may end up being
* discontiguous if some of the packets are short.
@@ -1426,7 +1498,6 @@ static int proc_do_submiturb(struct dev_state *ps, struct usbdevfs_urb *uurb,
}
kfree(isopkt);
isopkt = NULL;
- as->ps = ps;
as->userurb = arg;
if (is_in && uurb->buffer_length > 0)
as->userbuffer = uurb->buffer;
@@ -1487,6 +1558,8 @@ static int proc_do_submiturb(struct dev_state *ps, struct usbdevfs_urb *uurb,
return 0;
error:
+ if (usbm)
+ usbm->usb_use_count--;
kfree(isopkt);
kfree(dr);
if (as)
@@ -1969,6 +2042,64 @@ static int proc_disconnect_claim(struct dev_state *ps, void __user *arg)
return claimintf(ps, dc.interface);
}
+static int proc_release_memory(struct dev_state *ps, void __user *arg)
+{
+ struct usbdevfs_memory m;
+ struct usb_memory *usbm;
+ unsigned long flags;
+
+ if (copy_from_user(&m, arg, sizeof(m)))
+ return -EFAULT;
+
+ spin_lock_irqsave(&ps->lock, flags);
+ list_for_each_entry(usbm, &ps->memory_list, memlist) {
+ if ((usbm->vm_start == (unsigned long)m.buffer || usbm->offset == m.offset) &&
+ usbm->usb_use_count == 0 && usbm->vma_use_count == 0) {
+
+ list_del_init(&usbm->memlist);
+ spin_unlock_irqrestore(&ps->lock, flags);
+ free_pages_exact(usbm->mem, usbm->size);
+ kfree(usbm);
+ return 0;
+ }
+ }
+ spin_unlock_irqrestore(&ps->lock, flags);
+ return -EBUSY;
+}
+
+static int proc_alloc_memory(struct dev_state *ps, void __user *arg)
+{
+ struct usbdevfs_memory m;
+ struct usb_memory *usbmem;
+ void *mem;
+ unsigned long flags;
+
+ if (copy_from_user(&m, arg, sizeof(m)))
+ return -EFAULT;
+
+ mem = alloc_pages_exact(m.size, GFP_KERNEL | GFP_DMA32);
+ if (!mem)
+ return -ENOMEM;
+
+ usbmem = kzalloc(sizeof(struct usb_memory), GFP_KERNEL);
+ if (!usbmem) {
+ free_pages_exact(mem, m.size);
+ return -ENOMEM;
+ }
+ memset(mem, 0x0, (PAGE_SIZE<<get_order(m.size)));
+ usbmem->mem = mem;
+ m.offset = usbmem->offset = virt_to_phys(mem);
+ usbmem->size = m.size;
+ spin_lock_irqsave(&ps->lock, flags);
+ list_add_tail(&usbmem->memlist, &ps->memory_list);
+ spin_unlock_irqrestore(&ps->lock, flags);
+
+ if (copy_to_user(arg, &m, sizeof(m)))
+ return -EFAULT;
+
+ return 0;
+}
+
/*
* NOTE: All requests here that have interface numbers as parameters
* are assuming that somehow the configuration has been prevented from
@@ -2145,7 +2276,16 @@ static long usbdev_do_ioctl(struct file *file, unsigned int cmd,
case USBDEVFS_DISCONNECT_CLAIM:
ret = proc_disconnect_claim(ps, p);
break;
+ case USBDEVFS_ALLOC_MEMORY:
+ snoop(&dev->dev, "%s: ALLOC_MEMORY\n", __func__);
+ ret = proc_alloc_memory(ps, p);
+ break;
+ case USBDEVFS_RELEASE_MEMORY:
+ snoop(&dev->dev, "%s: RELEASE_MEMORY\n", __func__);
+ ret = proc_release_memory(ps, p);
+ break;
}
+
usb_unlock_device(dev);
if (ret >= 0)
inode->i_atime = CURRENT_TIME;
@@ -2162,6 +2302,60 @@ static long usbdev_ioctl(struct file *file, unsigned int cmd,
return ret;
}
+static void usbdev_vm_open(struct vm_area_struct *vma)
+{
+ struct usb_memory *usbm = vma->vm_private_data;
+ usbm->vma_use_count++;
+}
+
+static void usbdev_vm_close(struct vm_area_struct *vma)
+{
+ struct usb_memory *usbm = vma->vm_private_data;
+ usbm->vma_use_count--;
+}
+
+
+struct vm_operations_struct usbdev_vm_ops = {
+ .open = usbdev_vm_open,
+ .close = usbdev_vm_close
+};
+
+static int usbdev_mmap(struct file *file, struct vm_area_struct *vma)
+{
+ struct usb_memory *usbm = NULL, *usbm_iter = NULL;
+ struct dev_state *ps = file->private_data;
+ int size = vma->vm_end - vma->vm_start;
+ unsigned long flags;
+ spin_lock_irqsave(&ps->lock, flags);
+ list_for_each_entry(usbm_iter, &ps->memory_list, memlist) {
+
+ if (usbm_iter->offset == (vma->vm_pgoff<<PAGE_SHIFT) &&
+ size <= (PAGE_SIZE<<get_order(usbm_iter->size))) {
+ usbm = usbm_iter;
+ usbm->vm_start = vma->vm_start;
+ break;
+ }
+ }
+ spin_unlock_irqrestore(&ps->lock, flags);
+
+ if (usbm == NULL)
+ return -EINVAL;
+
+ vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+
+ if (remap_pfn_range(vma, vma->vm_start, virt_to_phys(usbm->mem) >> PAGE_SHIFT,
+ size,
+ vma->vm_page_prot) < 0)
+ return -EAGAIN;
+
+ vma->vm_flags |= VM_IO;
+ vma->vm_flags |= (VM_DONTEXPAND | VM_DONTDUMP);
+ vma->vm_ops = &usbdev_vm_ops;
+ vma->vm_private_data = usbm;
+ usbdev_vm_open(vma);
+ return 0;
+}
+
#ifdef CONFIG_COMPAT
static long usbdev_compat_ioctl(struct file *file, unsigned int cmd,
unsigned long arg)
@@ -2198,6 +2392,7 @@ const struct file_operations usbdev_file_operations = {
#ifdef CONFIG_COMPAT
.compat_ioctl = usbdev_compat_ioctl,
#endif
+ .mmap = usbdev_mmap,
.open = usbdev_open,
.release = usbdev_release,
};
diff --git a/include/uapi/linux/usbdevice_fs.h b/include/uapi/linux/usbdevice_fs.h
index 0c65e4b..1e63eec 100644
--- a/include/uapi/linux/usbdevice_fs.h
+++ b/include/uapi/linux/usbdevice_fs.h
@@ -144,6 +144,11 @@ struct usbdevfs_disconnect_claim {
char driver[USBDEVFS_MAXDRIVERNAME + 1];
};
+struct usbdevfs_memory {
+ u32 size;
+ u32 offset;
+ void __user *buffer;
+};
#define USBDEVFS_CONTROL _IOWR('U', 0, struct usbdevfs_ctrltransfer)
#define USBDEVFS_CONTROL32 _IOWR('U', 0, struct usbdevfs_ctrltransfer32)
@@ -176,5 +181,7 @@ struct usbdevfs_disconnect_claim {
#define USBDEVFS_RELEASE_PORT _IOR('U', 25, unsigned int)
#define USBDEVFS_GET_CAPABILITIES _IOR('U', 26, __u32)
#define USBDEVFS_DISCONNECT_CLAIM _IOR('U', 27, struct usbdevfs_disconnect_claim)
+#define USBDEVFS_ALLOC_MEMORY _IOWR('U', 28, struct usbdevfs_memory)
+#define USBDEVFS_RELEASE_MEMORY _IOW('U', 29, struct usbdevfs_memory)
#endif /* _UAPI_LINUX_USBDEVICE_FS_H */
--
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
