USB: add binary API to usbmon

This patch adds a new, "binary" API in addition to the old, text API usbmon
had before. The new API allows for less CPU use, and it allows to capture
all data from a packet where old API only captured 32 bytes at most. There
are some limitations and conditions to this, e.g. in case someone constructs
a URB with 1GB of data, it's not likely to be captured, because even the
huge buffers of the new reader are finite. Nonetheless, I expect this new
capability to capture all data for all real life scenarios.

The downside is, a special user mode application is required where cat(1)
worked before. I have sample code at http://people.redhat.com/zaitcev/linux/
and Paolo Abeni is working on patching libpcap.

This patch was initially written by Paolo and later I tweaked it, and
we had a little back-and-forth. So this is a jointly authored patch, but
I am submitting this I am responsible for the bugs.
Signed-off-by: NPaolo Abeni <paolo.abeni@email.it>
Signed-off-by: NPete Zaitcev <zaitcev@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Documentation/usb/usbmon.txt

@@ -77,7 +77,7 @@ that the file size is not excessive for your favourite editor.
 
 The '1t' type data consists of a stream of events, such as URB submission,
 URB callback, submission error. Every event is a text line, which consists
-of whitespace separated words. The number of position of words may depend
+of whitespace separated words. The number or position of words may depend
 on the event type, but there is a set of words, common for all types.
 
 Here is the list of words, from left to right:
@@ -170,4 +170,152 @@ dd65f0e8 4128379808 C Bo:005:02 0 31 >
 
 * Raw binary format and API
 
-TBD
+The overall architecture of the API is about the same as the one above,
+only the events are delivered in binary format. Each event is sent in
+the following structure (its name is made up, so that we can refer to it):
+
+struct usbmon_packet {
+	u64 id;			/*  0: URB ID - from submission to callback */
+	unsigned char type;	/*  8: Same as text; extensible. */
+	unsigned char xfer_type; /*    ISO (0), Intr, Control, Bulk (3) */
+	unsigned char epnum;	/*     Endpoint number and transfer direction */
+	unsigned char devnum;	/*     Device address */
+	u16 busnum;		/* 12: Bus number */
+	char flag_setup;	/* 14: Same as text */
+	char flag_data;		/* 15: Same as text; Binary zero is OK. */
+	s64 ts_sec;		/* 16: gettimeofday */
+	s32 ts_usec;		/* 24: gettimeofday */
+	int status;		/* 28: */
+	unsigned int length;	/* 32: Length of data (submitted or actual) */
+	unsigned int len_cap;	/* 36: Delivered length */
+	unsigned char setup[8];	/* 40: Only for Control 'S' */
+};				/* 48 bytes total */
+
+These events can be received from a character device by reading with read(2),
+with an ioctl(2), or by accessing the buffer with mmap.
+
+The character device is usually called /dev/usbmonN, where N is the USB bus
+number. Number zero (/dev/usbmon0) is special and means "all buses".
+However, this feature is not implemented yet. Note that specific naming
+policy is set by your Linux distribution.
+
+If you create /dev/usbmon0 by hand, make sure that it is owned by root
+and has mode 0600. Otherwise, unpriviledged users will be able to snoop
+keyboard traffic.
+
+The following ioctl calls are available, with MON_IOC_MAGIC 0x92:
+
+ MON_IOCQ_URB_LEN, defined as _IO(MON_IOC_MAGIC, 1)
+
+This call returns the length of data in the next event. Note that majority of
+events contain no data, so if this call returns zero, it does not mean that
+no events are available.
+
+ MON_IOCG_STATS, defined as _IOR(MON_IOC_MAGIC, 3, struct mon_bin_stats)
+
+The argument is a pointer to the following structure:
+
+struct mon_bin_stats {
+	u32 queued;
+	u32 dropped;
+};
+
+The member "queued" refers to the number of events currently queued in the
+buffer (and not to the number of events processed since the last reset).
+
+The member "dropped" is the number of events lost since the last call
+to MON_IOCG_STATS.
+
+ MON_IOCT_RING_SIZE, defined as _IO(MON_IOC_MAGIC, 4)
+
+This call sets the buffer size. The argument is the size in bytes.
+The size may be rounded down to the next chunk (or page). If the requested
+size is out of [unspecified] bounds for this kernel, the call fails with
+-EINVAL.
+
+ MON_IOCQ_RING_SIZE, defined as _IO(MON_IOC_MAGIC, 5)
+
+This call returns the current size of the buffer in bytes.
+
+ MON_IOCX_GET, defined as _IOW(MON_IOC_MAGIC, 6, struct mon_get_arg)
+
+This call waits for events to arrive if none were in the kernel buffer,
+then returns the first event. Its argument is a pointer to the following
+structure:
+
+struct mon_get_arg {
+	struct usbmon_packet *hdr;
+	void *data;
+	size_t alloc;		/* Length of data (can be zero) */
+};
+
+Before the call, hdr, data, and alloc should be filled. Upon return, the area
+pointed by hdr contains the next event structure, and the data buffer contains
+the data, if any. The event is removed from the kernel buffer.
+
+ MON_IOCX_MFETCH, defined as _IOWR(MON_IOC_MAGIC, 7, struct mon_mfetch_arg)
+
+This ioctl is primarily used when the application accesses the buffer
+with mmap(2). Its argument is a pointer to the following structure:
+
+struct mon_mfetch_arg {
+	uint32_t *offvec;	/* Vector of events fetched */
+	uint32_t nfetch;	/* Number of events to fetch (out: fetched) */
+	uint32_t nflush;	/* Number of events to flush */
+};
+
+The ioctl operates in 3 stages.
+
+First, it removes and discards up to nflush events from the kernel buffer.
+The actual number of events discarded is returned in nflush.
+
+Second, it waits for an event to be present in the buffer, unless the pseudo-
+device is open with O_NONBLOCK.
+
+Third, it extracts up to nfetch offsets into the mmap buffer, and stores
+them into the offvec. The actual number of event offsets is stored into
+the nfetch.
+
+ MON_IOCH_MFLUSH, defined as _IO(MON_IOC_MAGIC, 8)
+
+This call removes a number of events from the kernel buffer. Its argument
+is the number of events to remove. If the buffer contains fewer events
+than requested, all events present are removed, and no error is reported.
+This works when no events are available too.
+
+ FIONBIO
+
+The ioctl FIONBIO may be implemented in the future, if there's a need.
+
+In addition to ioctl(2) and read(2), the special file of binary API can
+be polled with select(2) and poll(2). But lseek(2) does not work.
+
+* Memory-mapped access of the kernel buffer for the binary API
+
+The basic idea is simple:
+
+To prepare, map the buffer by getting the current size, then using mmap(2).
+Then, execute a loop similar to the one written in pseudo-code below:
+
+   struct mon_mfetch_arg fetch;
+   struct usbmon_packet *hdr;
+   int nflush = 0;
+   for (;;) {
+      fetch.offvec = vec; // Has N 32-bit words
+      fetch.nfetch = N;   // Or less than N
+      fetch.nflush = nflush;
+      ioctl(fd, MON_IOCX_MFETCH, &fetch);   // Process errors, too
+      nflush = fetch.nfetch;       // This many packets to flush when done
+      for (i = 0; i < nflush; i++) {
+         hdr = (struct ubsmon_packet *) &mmap_area[vec[i]];
+         if (hdr->type == '@')     // Filler packet
+            continue;
+         caddr_t data = &mmap_area[vec[i]] + 64;
+         process_packet(hdr, data);
+      }
+   }
+
+Thus, the main idea is to execute only one ioctl per N events.
+
+Although the buffer is circular, the returned headers and data do not cross
+the end of the buffer, so the above pseudo-code does not need any gathering.

drivers/usb/mon/Makefile

@@ -2,7 +2,7 @@
 # Makefile for USB Core files and filesystem
 #
 
-usbmon-objs	:= mon_main.o mon_stat.o mon_text.o mon_dma.o
+usbmon-objs	:= mon_main.o mon_stat.o mon_text.o mon_bin.o mon_dma.o
 
 # This does not use CONFIG_USB_MON because we want this to use a tristate.
 obj-$(CONFIG_USB)	+= usbmon.o

drivers/usb/mon/mon_bin.c

+/*
+ * The USB Monitor, inspired by Dave Harding's USBMon.
+ *
+ * This is a binary format reader.
+ *
+ * Copyright (C) 2006 Paolo Abeni (paolo.abeni@email.it)
+ * Copyright (C) 2006 Pete Zaitcev (zaitcev@redhat.com)
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/usb.h>
+#include <linux/poll.h>
+#include <linux/compat.h>
+#include <linux/mm.h>
+
+#include <asm/uaccess.h>
+
+#include "usb_mon.h"
+
+/*
+ * Defined by USB 2.0 clause 9.3, table 9.2.
+ */
+#define SETUP_LEN  8
+
+/* ioctl macros */
+#define MON_IOC_MAGIC 0x92
+
+#define MON_IOCQ_URB_LEN _IO(MON_IOC_MAGIC, 1)
+/* #2 used to be MON_IOCX_URB, removed before it got into Linus tree */
+#define MON_IOCG_STATS _IOR(MON_IOC_MAGIC, 3, struct mon_bin_stats)
+#define MON_IOCT_RING_SIZE _IO(MON_IOC_MAGIC, 4)
+#define MON_IOCQ_RING_SIZE _IO(MON_IOC_MAGIC, 5)
+#define MON_IOCX_GET   _IOW(MON_IOC_MAGIC, 6, struct mon_bin_get)
+#define MON_IOCX_MFETCH _IOWR(MON_IOC_MAGIC, 7, struct mon_bin_mfetch)
+#define MON_IOCH_MFLUSH _IO(MON_IOC_MAGIC, 8)
+#ifdef CONFIG_COMPAT
+#define MON_IOCX_GET32 _IOW(MON_IOC_MAGIC, 6, struct mon_bin_get32)
+#define MON_IOCX_MFETCH32 _IOWR(MON_IOC_MAGIC, 7, struct mon_bin_mfetch32)
+#endif
+
+/*
+ * Some architectures have enormous basic pages (16KB for ia64, 64KB for ppc).
+ * But it's all right. Just use a simple way to make sure the chunk is never
+ * smaller than a page.
+ *
+ * N.B. An application does not know our chunk size.
+ *
+ * Woops, get_zeroed_page() returns a single page. I guess we're stuck with
+ * page-sized chunks for the time being.
+ */
+#define CHUNK_SIZE   PAGE_SIZE
+#define CHUNK_ALIGN(x)   (((x)+CHUNK_SIZE-1) & ~(CHUNK_SIZE-1))
+
+/*
+ * The magic limit was calculated so that it allows the monitoring
+ * application to pick data once in two ticks. This way, another application,
+ * which presumably drives the bus, gets to hog CPU, yet we collect our data.
+ * If HZ is 100, a 480 mbit/s bus drives 614 KB every jiffy. USB has an
+ * enormous overhead built into the bus protocol, so we need about 1000 KB.
+ *
+ * This is still too much for most cases, where we just snoop a few
+ * descriptor fetches for enumeration. So, the default is a "reasonable"
+ * amount for systems with HZ=250 and incomplete bus saturation.
+ *
+ * XXX What about multi-megabyte URBs which take minutes to transfer?
+ */
+#define BUFF_MAX  CHUNK_ALIGN(1200*1024)
+#define BUFF_DFL   CHUNK_ALIGN(300*1024)
+#define BUFF_MIN     CHUNK_ALIGN(8*1024)
+
+/*
+ * The per-event API header (2 per URB).
+ *
+ * This structure is seen in userland as defined by the documentation.
+ */
+struct mon_bin_hdr {
+	u64 id;			/* URB ID - from submission to callback */
+	unsigned char type;	/* Same as in text API; extensible. */
+	unsigned char xfer_type;	/* ISO, Intr, Control, Bulk */
+	unsigned char epnum;	/* Endpoint number and transfer direction */
+	unsigned char devnum;	/* Device address */
+	unsigned short busnum;	/* Bus number */
+	char flag_setup;
+	char flag_data;
+	s64 ts_sec;		/* gettimeofday */
+	s32 ts_usec;		/* gettimeofday */
+	int status;
+	unsigned int len_urb;	/* Length of data (submitted or actual) */
+	unsigned int len_cap;	/* Delivered length */
+	unsigned char setup[SETUP_LEN];	/* Only for Control S-type */
+};
+
+/* per file statistic */
+struct mon_bin_stats {
+	u32 queued;
+	u32 dropped;
+};
+
+struct mon_bin_get {
+	struct mon_bin_hdr __user *hdr;	/* Only 48 bytes, not 64. */
+	void __user *data;
+	size_t alloc;		/* Length of data (can be zero) */
+};
+
+struct mon_bin_mfetch {
+	u32 __user *offvec;	/* Vector of events fetched */
+	u32 nfetch;		/* Number of events to fetch (out: fetched) */
+	u32 nflush;		/* Number of events to flush */
+};
+
+#ifdef CONFIG_COMPAT
+struct mon_bin_get32 {
+	u32 hdr32;
+	u32 data32;
+	u32 alloc32;
+};
+
+struct mon_bin_mfetch32 {
+        u32 offvec32;
+        u32 nfetch32;
+        u32 nflush32;
+};
+#endif
+
+/* Having these two values same prevents wrapping of the mon_bin_hdr */
+#define PKT_ALIGN   64
+#define PKT_SIZE    64
+
+/* max number of USB bus supported */
+#define MON_BIN_MAX_MINOR 128
+
+/*
+ * The buffer: map of used pages.
+ */
+struct mon_pgmap {
+	struct page *pg;
+	unsigned char *ptr;	/* XXX just use page_to_virt everywhere? */
+};
+
+/*
+ * This gets associated with an open file struct.
+ */
+struct mon_reader_bin {
+	/* The buffer: one per open. */
+	spinlock_t b_lock;		/* Protect b_cnt, b_in */
+	unsigned int b_size;		/* Current size of the buffer - bytes */
+	unsigned int b_cnt;		/* Bytes used */
+	unsigned int b_in, b_out;	/* Offsets into buffer - bytes */
+	unsigned int b_read;		/* Amount of read data in curr. pkt. */
+	struct mon_pgmap *b_vec;	/* The map array */
+	wait_queue_head_t b_wait;	/* Wait for data here */
+
+	struct mutex fetch_lock;	/* Protect b_read, b_out */
+	int mmap_active;
+
+	/* A list of these is needed for "bus 0". Some time later. */
+	struct mon_reader r;
+
+	/* Stats */
+	unsigned int cnt_lost;
+};
+
+static inline struct mon_bin_hdr *MON_OFF2HDR(const struct mon_reader_bin *rp,
+    unsigned int offset)
+{
+	return (struct mon_bin_hdr *)
+	    (rp->b_vec[offset / CHUNK_SIZE].ptr + offset % CHUNK_SIZE);
+}
+
+#define MON_RING_EMPTY(rp)	((rp)->b_cnt == 0)
+
+static dev_t mon_bin_dev0;
+static struct cdev mon_bin_cdev;
+
+static void mon_buff_area_fill(const struct mon_reader_bin *rp,
+    unsigned int offset, unsigned int size);
+static int mon_bin_wait_event(struct file *file, struct mon_reader_bin *rp);
+static int mon_alloc_buff(struct mon_pgmap *map, int npages);
+static void mon_free_buff(struct mon_pgmap *map, int npages);
+
+/*
+ * This is a "chunked memcpy". It does not manipulate any counters.
+ * But it returns the new offset for repeated application.
+ */
+unsigned int mon_copy_to_buff(const struct mon_reader_bin *this,
+    unsigned int off, const unsigned char *from, unsigned int length)
+{
+	unsigned int step_len;
+	unsigned char *buf;
+	unsigned int in_page;
+
+	while (length) {
+		/*
+		 * Determine step_len.
+		 */
+		step_len = length;
+		in_page = CHUNK_SIZE - (off & (CHUNK_SIZE-1));
+		if (in_page < step_len)
+			step_len = in_page;
+
+		/*
+		 * Copy data and advance pointers.
+		 */
+		buf = this->b_vec[off / CHUNK_SIZE].ptr + off % CHUNK_SIZE;
+		memcpy(buf, from, step_len);
+		if ((off += step_len) >= this->b_size) off = 0;
+		from += step_len;
+		length -= step_len;
+	}
+	return off;
+}
+
+/*
+ * This is a little worse than the above because it's "chunked copy_to_user".
+ * The return value is an error code, not an offset.
+ */
+static int copy_from_buf(const struct mon_reader_bin *this, unsigned int off,
+    char __user *to, int length)
+{
+	unsigned int step_len;
+	unsigned char *buf;
+	unsigned int in_page;
+
+	while (length) {
+		/*
+		 * Determine step_len.
+		 */
+		step_len = length;
+		in_page = CHUNK_SIZE - (off & (CHUNK_SIZE-1));
+		if (in_page < step_len)
+			step_len = in_page;
+
+		/*
+		 * Copy data and advance pointers.
+		 */
+		buf = this->b_vec[off / CHUNK_SIZE].ptr + off % CHUNK_SIZE;
+		if (copy_to_user(to, buf, step_len))
+			return -EINVAL;
+		if ((off += step_len) >= this->b_size) off = 0;
+		to += step_len;
+		length -= step_len;
+	}
+	return 0;
+}
+
+/*
+ * Allocate an (aligned) area in the buffer.
+ * This is called under b_lock.
+ * Returns ~0 on failure.
+ */
+static unsigned int mon_buff_area_alloc(struct mon_reader_bin *rp,
+    unsigned int size)
+{
+	unsigned int offset;
+
+	size = (size + PKT_ALIGN-1) & ~(PKT_ALIGN-1);
+	if (rp->b_cnt + size > rp->b_size)
+		return ~0;
+	offset = rp->b_in;
+	rp->b_cnt += size;
+	if ((rp->b_in += size) >= rp->b_size)
+		rp->b_in -= rp->b_size;
+	return offset;
+}
+
+/*
+ * This is the same thing as mon_buff_area_alloc, only it does not allow
+ * buffers to wrap. This is needed by applications which pass references
+ * into mmap-ed buffers up their stacks (libpcap can do that).
+ *
+ * Currently, we always have the header stuck with the data, although
+ * it is not strictly speaking necessary.
+ *
+ * When a buffer would wrap, we place a filler packet to mark the space.
+ */
+static unsigned int mon_buff_area_alloc_contiguous(struct mon_reader_bin *rp,
+    unsigned int size)
+{
+	unsigned int offset;
+	unsigned int fill_size;
+
+	size = (size + PKT_ALIGN-1) & ~(PKT_ALIGN-1);
+	if (rp->b_cnt + size > rp->b_size)
+		return ~0;
+	if (rp->b_in + size > rp->b_size) {
+		/*
+		 * This would wrap. Find if we still have space after
+		 * skipping to the end of the buffer. If we do, place
+		 * a filler packet and allocate a new packet.
+		 */
+		fill_size = rp->b_size - rp->b_in;
+		if (rp->b_cnt + size + fill_size > rp->b_size)
+			return ~0;
+		mon_buff_area_fill(rp, rp->b_in, fill_size);
+
+		offset = 0;
+		rp->b_in = size;
+		rp->b_cnt += size + fill_size;
+	} else if (rp->b_in + size == rp->b_size) {
+		offset = rp->b_in;
+		rp->b_in = 0;
+		rp->b_cnt += size;
+	} else {
+		offset = rp->b_in;
+		rp->b_in += size;
+		rp->b_cnt += size;
+	}
+	return offset;
+}
+
+/*
+ * Return a few (kilo-)bytes to the head of the buffer.
+ * This is used if a DMA fetch fails.
+ */
+static void mon_buff_area_shrink(struct mon_reader_bin *rp, unsigned int size)
+{
+
+	size = (size + PKT_ALIGN-1) & ~(PKT_ALIGN-1);
+	rp->b_cnt -= size;
+	if (rp->b_in < size)
+		rp->b_in += rp->b_size;
+	rp->b_in -= size;
+}
+
+/*
+ * This has to be called under both b_lock and fetch_lock, because
+ * it accesses both b_cnt and b_out.
+ */
+static void mon_buff_area_free(struct mon_reader_bin *rp, unsigned int size)
+{
+
+	size = (size + PKT_ALIGN-1) & ~(PKT_ALIGN-1);
+	rp->b_cnt -= size;
+	if ((rp->b_out += size) >= rp->b_size)
+		rp->b_out -= rp->b_size;
+}
+
+static void mon_buff_area_fill(const struct mon_reader_bin *rp,
+    unsigned int offset, unsigned int size)
+{
+	struct mon_bin_hdr *ep;
+
+	ep = MON_OFF2HDR(rp, offset);
+	memset(ep, 0, PKT_SIZE);
+	ep->type = '@';
+	ep->len_cap = size - PKT_SIZE;
+}
+
+static inline char mon_bin_get_setup(unsigned char *setupb,
+    const struct urb *urb, char ev_type)
+{
+
+	if (!usb_pipecontrol(urb->pipe) || ev_type != 'S')
+		return '-';
+
+	if (urb->transfer_flags & URB_NO_SETUP_DMA_MAP)
+		return mon_dmapeek(setupb, urb->setup_dma, SETUP_LEN);
+	if (urb->setup_packet == NULL)
+		return 'Z';
+
+	memcpy(setupb, urb->setup_packet, SETUP_LEN);
+	return 0;
+}
+
+static char mon_bin_get_data(const struct mon_reader_bin *rp,
+    unsigned int offset, struct urb *urb, unsigned int length)
+{
+
+	if (urb->transfer_flags & URB_NO_TRANSFER_DMA_MAP) {
+		mon_dmapeek_vec(rp, offset, urb->transfer_dma, length);
+		return 0;
+	}
+
+	if (urb->transfer_buffer == NULL)
+		return 'Z';
+
+	mon_copy_to_buff(rp, offset, urb->transfer_buffer, length);
+	return 0;
+}
+
+static void mon_bin_event(struct mon_reader_bin *rp, struct urb *urb,
+    char ev_type)
+{
+	unsigned long flags;
+	struct timeval ts;
+	unsigned int urb_length;
+	unsigned int offset;
+	unsigned int length;
+	struct mon_bin_hdr *ep;
+	char data_tag = 0;
+
+	do_gettimeofday(&ts);
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+
+	/*
+	 * Find the maximum allowable length, then allocate space.
+	 */
+	urb_length = (ev_type == 'S') ?
+	    urb->transfer_buffer_length : urb->actual_length;
+	length = urb_length;
+
+	if (length >= rp->b_size/5)
+		length = rp->b_size/5;
+
+	if (usb_pipein(urb->pipe)) {
+		if (ev_type == 'S') {
+			length = 0;
+			data_tag = '<';
+		}
+	} else {
+		if (ev_type == 'C') {
+			length = 0;
+			data_tag = '>';
+		}
+	}
+
+	if (rp->mmap_active)
+		offset = mon_buff_area_alloc_contiguous(rp, length + PKT_SIZE);
+	else
+		offset = mon_buff_area_alloc(rp, length + PKT_SIZE);
+	if (offset == ~0) {
+		rp->cnt_lost++;
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+		return;
+	}
+
+	ep = MON_OFF2HDR(rp, offset);
+	if ((offset += PKT_SIZE) >= rp->b_size) offset = 0;
+
+	/*
+	 * Fill the allocated area.
+	 */
+	memset(ep, 0, PKT_SIZE);
+	ep->type = ev_type;
+	ep->xfer_type = usb_pipetype(urb->pipe);
+	/* We use the fact that usb_pipein() returns 0x80 */
+	ep->epnum = usb_pipeendpoint(urb->pipe) | usb_pipein(urb->pipe);
+	ep->devnum = usb_pipedevice(urb->pipe);
+	ep->busnum = rp->r.m_bus->u_bus->busnum;
+	ep->id = (unsigned long) urb;
+	ep->ts_sec = ts.tv_sec;
+	ep->ts_usec = ts.tv_usec;
+	ep->status = urb->status;
+	ep->len_urb = urb_length;
+	ep->len_cap = length;
+
+	ep->flag_setup = mon_bin_get_setup(ep->setup, urb, ev_type);
+	if (length != 0) {
+		ep->flag_data = mon_bin_get_data(rp, offset, urb, length);
+		if (ep->flag_data != 0) {	/* Yes, it's 0x00, not '0' */
+			ep->len_cap = 0;
+			mon_buff_area_shrink(rp, length);
+		}
+	} else {
+		ep->flag_data = data_tag;
+	}
+
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+
+	wake_up(&rp->b_wait);
+}
+
+static void mon_bin_submit(void *data, struct urb *urb)
+{
+	struct mon_reader_bin *rp = data;
+	mon_bin_event(rp, urb, 'S');
+}
+
+static void mon_bin_complete(void *data, struct urb *urb)
+{
+	struct mon_reader_bin *rp = data;
+	mon_bin_event(rp, urb, 'C');
+}
+
+static void mon_bin_error(void *data, struct urb *urb, int error)
+{
+	struct mon_reader_bin *rp = data;
+	unsigned long flags;
+	unsigned int offset;
+	struct mon_bin_hdr *ep;
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+
+	offset = mon_buff_area_alloc(rp, PKT_SIZE);
+	if (offset == ~0) {
+		/* Not incrementing cnt_lost. Just because. */
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+		return;
+	}
+
+	ep = MON_OFF2HDR(rp, offset);
+
+	memset(ep, 0, PKT_SIZE);
+	ep->type = 'E';
+	ep->xfer_type = usb_pipetype(urb->pipe);
+	/* We use the fact that usb_pipein() returns 0x80 */
+	ep->epnum = usb_pipeendpoint(urb->pipe) | usb_pipein(urb->pipe);
+	ep->devnum = usb_pipedevice(urb->pipe);
+	ep->busnum = rp->r.m_bus->u_bus->busnum;
+	ep->id = (unsigned long) urb;
+	ep->status = error;
+
+	ep->flag_setup = '-';
+	ep->flag_data = 'E';
+
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+
+	wake_up(&rp->b_wait);
+}
+
+static int mon_bin_open(struct inode *inode, struct file *file)
+{
+	struct mon_bus *mbus;
+	struct usb_bus *ubus;
+	struct mon_reader_bin *rp;
+	size_t size;
+	int rc;
+
+	mutex_lock(&mon_lock);
+	if ((mbus = mon_bus_lookup(iminor(inode))) == NULL) {
+		mutex_unlock(&mon_lock);
+		return -ENODEV;
+	}
+	if ((ubus = mbus->u_bus) == NULL) {
+		printk(KERN_ERR TAG ": consistency error on open\n");
+		mutex_unlock(&mon_lock);
+		return -ENODEV;
+	}
+
+	rp = kzalloc(sizeof(struct mon_reader_bin), GFP_KERNEL);
+	if (rp == NULL) {
+		rc = -ENOMEM;
+		goto err_alloc;
+	}
+	spin_lock_init(&rp->b_lock);
+	init_waitqueue_head(&rp->b_wait);
+	mutex_init(&rp->fetch_lock);
+
+	rp->b_size = BUFF_DFL;
+
+	size = sizeof(struct mon_pgmap) * (rp->b_size/CHUNK_SIZE);
+	if ((rp->b_vec = kzalloc(size, GFP_KERNEL)) == NULL) {
+		rc = -ENOMEM;
+		goto err_allocvec;
+	}
+
+	if ((rc = mon_alloc_buff(rp->b_vec, rp->b_size/CHUNK_SIZE)) < 0)
+		goto err_allocbuff;
+
+	rp->r.m_bus = mbus;
+	rp->r.r_data = rp;
+	rp->r.rnf_submit = mon_bin_submit;
+	rp->r.rnf_error = mon_bin_error;
+	rp->r.rnf_complete = mon_bin_complete;
+
+	mon_reader_add(mbus, &rp->r);
+
+	file->private_data = rp;
+	mutex_unlock(&mon_lock);
+	return 0;
+
+err_allocbuff:
+	kfree(rp->b_vec);
+err_allocvec:
+	kfree(rp);
+err_alloc:
+	mutex_unlock(&mon_lock);
+	return rc;
+}
+
+/*
+ * Extract an event from buffer and copy it to user space.
+ * Wait if there is no event ready.
+ * Returns zero or error.
+ */
+static int mon_bin_get_event(struct file *file, struct mon_reader_bin *rp,
+    struct mon_bin_hdr __user *hdr, void __user *data, unsigned int nbytes)
+{
+	unsigned long flags;
+	struct mon_bin_hdr *ep;
+	size_t step_len;
+	unsigned int offset;
+	int rc;
+
+	mutex_lock(&rp->fetch_lock);
+
+	if ((rc = mon_bin_wait_event(file, rp)) < 0) {
+		mutex_unlock(&rp->fetch_lock);
+		return rc;
+	}
+
+	ep = MON_OFF2HDR(rp, rp->b_out);
+
+	if (copy_to_user(hdr, ep, sizeof(struct mon_bin_hdr))) {
+		mutex_unlock(&rp->fetch_lock);
+		return -EFAULT;
+	}
+
+	step_len = min(ep->len_cap, nbytes);
+	if ((offset = rp->b_out + PKT_SIZE) >= rp->b_size) offset = 0;
+
+	if (copy_from_buf(rp, offset, data, step_len)) {
+		mutex_unlock(&rp->fetch_lock);
+		return -EFAULT;
+	}
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+	mon_buff_area_free(rp, PKT_SIZE + ep->len_cap);
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+	rp->b_read = 0;
+
+	mutex_unlock(&rp->fetch_lock);
+	return 0;
+}
+
+static int mon_bin_release(struct inode *inode, struct file *file)
+{
+	struct mon_reader_bin *rp = file->private_data;
+	struct mon_bus* mbus = rp->r.m_bus;
+
+	mutex_lock(&mon_lock);
+
+	if (mbus->nreaders <= 0) {
+		printk(KERN_ERR TAG ": consistency error on close\n");
+		mutex_unlock(&mon_lock);
+		return 0;
+	}
+	mon_reader_del(mbus, &rp->r);
+
+	mon_free_buff(rp->b_vec, rp->b_size/CHUNK_SIZE);
+	kfree(rp->b_vec);
+	kfree(rp);
+
+	mutex_unlock(&mon_lock);
+	return 0;
+}
+
+static ssize_t mon_bin_read(struct file *file, char __user *buf,
+    size_t nbytes, loff_t *ppos)
+{
+	struct mon_reader_bin *rp = file->private_data;
+	unsigned long flags;
+	struct mon_bin_hdr *ep;
+	unsigned int offset;
+	size_t step_len;
+	char *ptr;
+	ssize_t done = 0;
+	int rc;
+
+	mutex_lock(&rp->fetch_lock);
+
+	if ((rc = mon_bin_wait_event(file, rp)) < 0) {
+		mutex_unlock(&rp->fetch_lock);
+		return rc;
+	}
+
+	ep = MON_OFF2HDR(rp, rp->b_out);
+
+	if (rp->b_read < sizeof(struct mon_bin_hdr)) {
+		step_len = min(nbytes, sizeof(struct mon_bin_hdr) - rp->b_read);
+		ptr = ((char *)ep) + rp->b_read;
+		if (step_len && copy_to_user(buf, ptr, step_len)) {
+			mutex_unlock(&rp->fetch_lock);
+			return -EFAULT;
+		}
+		nbytes -= step_len;
+		buf += step_len;
+		rp->b_read += step_len;
+		done += step_len;
+	}
+
+	if (rp->b_read >= sizeof(struct mon_bin_hdr)) {
+		step_len = min(nbytes, (size_t)ep->len_cap);
+		offset = rp->b_out + PKT_SIZE;
+		offset += rp->b_read - sizeof(struct mon_bin_hdr);
+		if (offset >= rp->b_size)
+			offset -= rp->b_size;
+		if (copy_from_buf(rp, offset, buf, step_len)) {
+			mutex_unlock(&rp->fetch_lock);
+			return -EFAULT;
+		}
+		nbytes -= step_len;
+		buf += step_len;
+		rp->b_read += step_len;
+		done += step_len;
+	}
+
+	/*
+	 * Check if whole packet was read, and if so, jump to the next one.
+	 */
+	if (rp->b_read >= sizeof(struct mon_bin_hdr) + ep->len_cap) {
+		spin_lock_irqsave(&rp->b_lock, flags);
+		mon_buff_area_free(rp, PKT_SIZE + ep->len_cap);
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+		rp->b_read = 0;
+	}
+
+	mutex_unlock(&rp->fetch_lock);
+	return done;
+}
+
+/*
+ * Remove at most nevents from chunked buffer.
+ * Returns the number of removed events.
+ */
+static int mon_bin_flush(struct mon_reader_bin *rp, unsigned nevents)
+{
+	unsigned long flags;
+	struct mon_bin_hdr *ep;
+	int i;
+
+	mutex_lock(&rp->fetch_lock);
+	spin_lock_irqsave(&rp->b_lock, flags);
+	for (i = 0; i < nevents; ++i) {
+		if (MON_RING_EMPTY(rp))
+			break;
+
+		ep = MON_OFF2HDR(rp, rp->b_out);
+		mon_buff_area_free(rp, PKT_SIZE + ep->len_cap);
+	}
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+	rp->b_read = 0;
+	mutex_unlock(&rp->fetch_lock);
+	return i;
+}
+
+/*
+ * Fetch at most max event offsets into the buffer and put them into vec.
+ * The events are usually freed later with mon_bin_flush.
+ * Return the effective number of events fetched.
+ */
+static int mon_bin_fetch(struct file *file, struct mon_reader_bin *rp,
+    u32 __user *vec, unsigned int max)
+{
+	unsigned int cur_out;
+	unsigned int bytes, avail;
+	unsigned int size;
+	unsigned int nevents;
+	struct mon_bin_hdr *ep;
+	unsigned long flags;
+	int rc;
+
+	mutex_lock(&rp->fetch_lock);
+
+	if ((rc = mon_bin_wait_event(file, rp)) < 0) {
+		mutex_unlock(&rp->fetch_lock);
+		return rc;
+	}
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+	avail = rp->b_cnt;
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+
+	cur_out = rp->b_out;
+	nevents = 0;
+	bytes = 0;
+	while (bytes < avail) {
+		if (nevents >= max)
+			break;
+
+		ep = MON_OFF2HDR(rp, cur_out);
+		if (put_user(cur_out, &vec[nevents])) {
+			mutex_unlock(&rp->fetch_lock);
+			return -EFAULT;
+		}
+
+		nevents++;
+		size = ep->len_cap + PKT_SIZE;
+		size = (size + PKT_ALIGN-1) & ~(PKT_ALIGN-1);
+		if ((cur_out += size) >= rp->b_size)
+			cur_out -= rp->b_size;
+		bytes += size;
+	}
+
+	mutex_unlock(&rp->fetch_lock);
+	return nevents;
+}
+
+/*
+ * Count events. This is almost the same as the above mon_bin_fetch,
+ * only we do not store offsets into user vector, and we have no limit.
+ */
+static int mon_bin_queued(struct mon_reader_bin *rp)
+{
+	unsigned int cur_out;
+	unsigned int bytes, avail;
+	unsigned int size;
+	unsigned int nevents;
+	struct mon_bin_hdr *ep;
+	unsigned long flags;
+
+	mutex_lock(&rp->fetch_lock);
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+	avail = rp->b_cnt;
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+
+	cur_out = rp->b_out;
+	nevents = 0;
+	bytes = 0;
+	while (bytes < avail) {
+		ep = MON_OFF2HDR(rp, cur_out);
+
+		nevents++;
+		size = ep->len_cap + PKT_SIZE;
+		size = (size + PKT_ALIGN-1) & ~(PKT_ALIGN-1);
+		if ((cur_out += size) >= rp->b_size)
+			cur_out -= rp->b_size;
+		bytes += size;
+	}
+
+	mutex_unlock(&rp->fetch_lock);
+	return nevents;
+}
+
+/*
+ */
+static int mon_bin_ioctl(struct inode *inode, struct file *file,
+    unsigned int cmd, unsigned long arg)
+{
+	struct mon_reader_bin *rp = file->private_data;
+	// struct mon_bus* mbus = rp->r.m_bus;
+	int ret = 0;
+	struct mon_bin_hdr *ep;
+	unsigned long flags;
+
+	switch (cmd) {
+
+	case MON_IOCQ_URB_LEN:
+		/*
+		 * N.B. This only returns the size of data, without the header.
+		 */
+		spin_lock_irqsave(&rp->b_lock, flags);
+		if (!MON_RING_EMPTY(rp)) {
+			ep = MON_OFF2HDR(rp, rp->b_out);
+			ret = ep->len_cap;
+		}
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+		break;
+
+	case MON_IOCQ_RING_SIZE:
+		ret = rp->b_size;
+		break;
+
+	case MON_IOCT_RING_SIZE:
+		/*
+		 * Changing the buffer size will flush it's contents; the new
+		 * buffer is allocated before releasing the old one to be sure
+		 * the device will stay functional also in case of memory
+		 * pressure.
+		 */
+		{
+		int size;
+		struct mon_pgmap *vec;
+
+		if (arg < BUFF_MIN || arg > BUFF_MAX)
+			return -EINVAL;
+
+		size = CHUNK_ALIGN(arg);
+		if ((vec = kzalloc(sizeof(struct mon_pgmap) * (size/CHUNK_SIZE),
+		    GFP_KERNEL)) == NULL) {
+			ret = -ENOMEM;
+			break;
+		}
+
+		ret = mon_alloc_buff(vec, size/CHUNK_SIZE);
+		if (ret < 0) {
+			kfree(vec);
+			break;
+		}
+
+		mutex_lock(&rp->fetch_lock);
+		spin_lock_irqsave(&rp->b_lock, flags);
+		mon_free_buff(rp->b_vec, size/CHUNK_SIZE);
+		kfree(rp->b_vec);
+		rp->b_vec  = vec;
+		rp->b_size = size;
+		rp->b_read = rp->b_in = rp->b_out = rp->b_cnt = 0;
+		rp->cnt_lost = 0;
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+		mutex_unlock(&rp->fetch_lock);
+		}
+		break;
+
+	case MON_IOCH_MFLUSH:
+		ret = mon_bin_flush(rp, arg);
+		break;
+
+	case MON_IOCX_GET:
+		{
+		struct mon_bin_get getb;
+
+		if (copy_from_user(&getb, (void __user *)arg,
+					    sizeof(struct mon_bin_get)))
+			return -EFAULT;
+
+		if (getb.alloc > 0x10000000)	/* Want to cast to u32 */
+			return -EINVAL;
+		ret = mon_bin_get_event(file, rp,
+			  getb.hdr, getb.data, (unsigned int)getb.alloc);
+		}
+		break;
+
+#ifdef CONFIG_COMPAT
+	case MON_IOCX_GET32: {
+		struct mon_bin_get32 getb;
+
+		if (copy_from_user(&getb, (void __user *)arg,
+					    sizeof(struct mon_bin_get32)))
+			return -EFAULT;
+
+		ret = mon_bin_get_event(file, rp,
+		    compat_ptr(getb.hdr32), compat_ptr(getb.data32),
+		    getb.alloc32);
+		}
+		break;
+#endif
+
+	case MON_IOCX_MFETCH:
+		{
+		struct mon_bin_mfetch mfetch;
+		struct mon_bin_mfetch __user *uptr;
+
+		uptr = (struct mon_bin_mfetch __user *)arg;
+
+		if (copy_from_user(&mfetch, uptr, sizeof(mfetch)))
+			return -EFAULT;
+
+		if (mfetch.nflush) {
+			ret = mon_bin_flush(rp, mfetch.nflush);
+			if (ret < 0)
+				return ret;
+			if (put_user(ret, &uptr->nflush))
+				return -EFAULT;
+		}
+		ret = mon_bin_fetch(file, rp, mfetch.offvec, mfetch.nfetch);
+		if (ret < 0)
+			return ret;
+		if (put_user(ret, &uptr->nfetch))
+			return -EFAULT;
+		ret = 0;
+		}
+		break;
+
+#ifdef CONFIG_COMPAT
+	case MON_IOCX_MFETCH32:
+		{
+		struct mon_bin_mfetch32 mfetch;
+		struct mon_bin_mfetch32 __user *uptr;
+
+		uptr = (struct mon_bin_mfetch32 __user *) compat_ptr(arg);
+
+		if (copy_from_user(&mfetch, uptr, sizeof(mfetch)))
+			return -EFAULT;
+
+		if (mfetch.nflush32) {
+			ret = mon_bin_flush(rp, mfetch.nflush32);
+			if (ret < 0)
+				return ret;
+			if (put_user(ret, &uptr->nflush32))
+				return -EFAULT;
+		}
+		ret = mon_bin_fetch(file, rp, compat_ptr(mfetch.offvec32),
+		    mfetch.nfetch32);
+		if (ret < 0)
+			return ret;
+		if (put_user(ret, &uptr->nfetch32))
+			return -EFAULT;
+		ret = 0;
+		}
+		break;
+#endif
+
+	case MON_IOCG_STATS: {
+		struct mon_bin_stats __user *sp;
+		unsigned int nevents;
+		unsigned int ndropped;
+
+		spin_lock_irqsave(&rp->b_lock, flags);
+		ndropped = rp->cnt_lost;
+		rp->cnt_lost = 0;
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+		nevents = mon_bin_queued(rp);
+
+		sp = (struct mon_bin_stats __user *)arg;
+		if (put_user(rp->cnt_lost, &sp->dropped))
+			return -EFAULT;
+		if (put_user(nevents, &sp->queued))
+			return -EFAULT;
+
+		}
+		break;
+
+	default:
+		return -ENOTTY;
+	}
+
+	return ret;
+}
+
+static unsigned int
+mon_bin_poll(struct file *file, struct poll_table_struct *wait)
+{
+	struct mon_reader_bin *rp = file->private_data;
+	unsigned int mask = 0;
+	unsigned long flags;
+
+	if (file->f_mode & FMODE_READ)
+		poll_wait(file, &rp->b_wait, wait);
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+	if (!MON_RING_EMPTY(rp))
+		mask |= POLLIN | POLLRDNORM;    /* readable */
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+	return mask;
+}
+
+/*
+ * open and close: just keep track of how many times the device is
+ * mapped, to use the proper memory allocation function.
+ */
+static void mon_bin_vma_open(struct vm_area_struct *vma)
+{
+	struct mon_reader_bin *rp = vma->vm_private_data;
+	rp->mmap_active++;
+}
+
+static void mon_bin_vma_close(struct vm_area_struct *vma)
+{
+	struct mon_reader_bin *rp = vma->vm_private_data;
+	rp->mmap_active--;
+}
+
+/*
+ * Map ring pages to user space.
+ */
+struct page *mon_bin_vma_nopage(struct vm_area_struct *vma,
+                                unsigned long address, int *type)
+{
+	struct mon_reader_bin *rp = vma->vm_private_data;
+	unsigned long offset, chunk_idx;
+	struct page *pageptr;
+
+	offset = (address - vma->vm_start) + (vma->vm_pgoff << PAGE_SHIFT);
+	if (offset >= rp->b_size)
+		return NOPAGE_SIGBUS;
+	chunk_idx = offset / CHUNK_SIZE;
+	pageptr = rp->b_vec[chunk_idx].pg;
+	get_page(pageptr);
+	if (type)
+		*type = VM_FAULT_MINOR;
+	return pageptr;
+}
+
+struct vm_operations_struct mon_bin_vm_ops = {
+	.open =     mon_bin_vma_open,
+	.close =    mon_bin_vma_close,
+	.nopage =   mon_bin_vma_nopage,
+};
+
+int mon_bin_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	/* don't do anything here: "nopage" will set up page table entries */
+	vma->vm_ops = &mon_bin_vm_ops;
+	vma->vm_flags |= VM_RESERVED;
+	vma->vm_private_data = filp->private_data;
+	mon_bin_vma_open(vma);
+	return 0;
+}
+
+struct file_operations mon_fops_binary = {
+	.owner =	THIS_MODULE,
+	.open =		mon_bin_open,
+	.llseek =	no_llseek,
+	.read =		mon_bin_read,
+	/* .write =	mon_text_write, */
+	.poll =		mon_bin_poll,
+	.ioctl =	mon_bin_ioctl,
+	.release =	mon_bin_release,
+};
+
+static int mon_bin_wait_event(struct file *file, struct mon_reader_bin *rp)
+{
+	DECLARE_WAITQUEUE(waita, current);
+	unsigned long flags;
+
+	add_wait_queue(&rp->b_wait, &waita);
+	set_current_state(TASK_INTERRUPTIBLE);
+
+	spin_lock_irqsave(&rp->b_lock, flags);
+	while (MON_RING_EMPTY(rp)) {
+		spin_unlock_irqrestore(&rp->b_lock, flags);
+
+		if (file->f_flags & O_NONBLOCK) {
+			set_current_state(TASK_RUNNING);
+			remove_wait_queue(&rp->b_wait, &waita);
+			return -EWOULDBLOCK; /* Same as EAGAIN in Linux */
+		}
+		schedule();
+		if (signal_pending(current)) {
+			remove_wait_queue(&rp->b_wait, &waita);
+			return -EINTR;
+		}
+		set_current_state(TASK_INTERRUPTIBLE);
+
+		spin_lock_irqsave(&rp->b_lock, flags);
+	}
+	spin_unlock_irqrestore(&rp->b_lock, flags);
+
+	set_current_state(TASK_RUNNING);
+	remove_wait_queue(&rp->b_wait, &waita);
+	return 0;
+}
+
+static int mon_alloc_buff(struct mon_pgmap *map, int npages)
+{
+	int n;
+	unsigned long vaddr;
+
+	for (n = 0; n < npages; n++) {
+		vaddr = get_zeroed_page(GFP_KERNEL);
+		if (vaddr == 0) {
+			while (n-- != 0)
+				free_page((unsigned long) map[n].ptr);
+			return -ENOMEM;
+		}
+		map[n].ptr = (unsigned char *) vaddr;
+		map[n].pg = virt_to_page(vaddr);
+	}
+	return 0;
+}
+
+static void mon_free_buff(struct mon_pgmap *map, int npages)
+{
+	int n;
+
+	for (n = 0; n < npages; n++)
+		free_page((unsigned long) map[n].ptr);
+}
+
+int __init mon_bin_init(void)
+{
+	int rc;
+
+	rc = alloc_chrdev_region(&mon_bin_dev0, 0, MON_BIN_MAX_MINOR, "usbmon");
+	if (rc < 0)
+		goto err_dev;
+
+	cdev_init(&mon_bin_cdev, &mon_fops_binary);
+	mon_bin_cdev.owner = THIS_MODULE;
+
+	rc = cdev_add(&mon_bin_cdev, mon_bin_dev0, MON_BIN_MAX_MINOR);
+	if (rc < 0)
+		goto err_add;
+
+	return 0;
+
+err_add:
+	unregister_chrdev_region(mon_bin_dev0, MON_BIN_MAX_MINOR);
+err_dev:
+	return rc;
+}
+
+void __exit mon_bin_exit(void)
+{
+	cdev_del(&mon_bin_cdev);
+	unregister_chrdev_region(mon_bin_dev0, MON_BIN_MAX_MINOR);
+}

drivers/usb/mon/mon_dma.c

@@ -48,6 +48,36 @@ char mon_dmapeek(unsigned char *dst, dma_addr_t dma_addr, int len)
 	local_irq_restore(flags);
 	return 0;
 }
+
+void mon_dmapeek_vec(const struct mon_reader_bin *rp,
+    unsigned int offset, dma_addr_t dma_addr, unsigned int length)
+{
+	unsigned long flags;
+	unsigned int step_len;
+	struct page *pg;
+	unsigned char *map;
+	unsigned long page_off, page_len;
+
+	local_irq_save(flags);
+	while (length) {
+		/* compute number of bytes we are going to copy in this page */
+		step_len = length;
+		page_off = dma_addr & (PAGE_SIZE-1);
+		page_len = PAGE_SIZE - page_off;
+		if (page_len < step_len)
+			step_len = page_len;
+
+		/* copy data and advance pointers */
+		pg = phys_to_page(dma_addr);
+		map = kmap_atomic(pg, KM_IRQ0);
+		offset = mon_copy_to_buff(rp, offset, map + page_off, step_len);
+		kunmap_atomic(map, KM_IRQ0);
+		dma_addr += step_len;
+		length -= step_len;
+	}
+	local_irq_restore(flags);
+}
+
 #endif /* __i386__ */
 
 #ifndef MON_HAS_UNMAP
@@ -55,4 +85,11 @@ char mon_dmapeek(unsigned char *dst, dma_addr_t dma_addr, int len)
 {
 	return 'D';
 }
-#endif
+
+void mon_dmapeek_vec(const struct mon_reader_bin *rp,
+    unsigned int offset, dma_addr_t dma_addr, unsigned int length)
+{
+	;
+}
+
+#endif /* MON_HAS_UNMAP */

drivers/usb/mon/mon_main.c

@@ -9,7 +9,6 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/usb.h>
-#include <linux/debugfs.h>
 #include <linux/smp_lock.h>
 #include <linux/notifier.h>
 #include <linux/mutex.h>
@@ -22,11 +21,10 @@ static void mon_complete(struct usb_bus *ubus, struct urb *urb);
 static void mon_stop(struct mon_bus *mbus);
 static void mon_dissolve(struct mon_bus *mbus, struct usb_bus *ubus);
 static void mon_bus_drop(struct kref *r);
-static void mon_bus_init(struct dentry *mondir, struct usb_bus *ubus);
+static void mon_bus_init(struct usb_bus *ubus);
 
 DEFINE_MUTEX(mon_lock);
 
-static struct dentry *mon_dir;		/* /dbg/usbmon */
 static LIST_HEAD(mon_buses);		/* All buses we know: struct mon_bus */
 
 /*
@@ -200,7 +198,7 @@ static void mon_stop(struct mon_bus *mbus)
  */
 static void mon_bus_add(struct usb_bus *ubus)
 {
-	mon_bus_init(mon_dir, ubus);
+	mon_bus_init(ubus);
 }
 
 /*
@@ -212,8 +210,8 @@ static void mon_bus_remove(struct usb_bus *ubus)
 
 	mutex_lock(&mon_lock);
 	list_del(&mbus->bus_link);
-	debugfs_remove(mbus->dent_t);
-	debugfs_remove(mbus->dent_s);
+	if (mbus->text_inited)
+		mon_text_del(mbus);
 
 	mon_dissolve(mbus, ubus);
 	kref_put(&mbus->ref, mon_bus_drop);
@@ -281,13 +279,9 @@ static void mon_bus_drop(struct kref *r)
  *  - refcount USB bus struct
  *  - link
  */
-static void mon_bus_init(struct dentry *mondir, struct usb_bus *ubus)
+static void mon_bus_init(struct usb_bus *ubus)
 {
-	struct dentry *d;
 	struct mon_bus *mbus;
-	enum { NAMESZ = 10 };
-	char name[NAMESZ];
-	int rc;
 
 	if ((mbus = kzalloc(sizeof(struct mon_bus), GFP_KERNEL)) == NULL)
 		goto err_alloc;
@@ -303,57 +297,54 @@ static void mon_bus_init(struct dentry *mondir, struct usb_bus *ubus)
 	ubus->mon_bus = mbus;
 	mbus->uses_dma = ubus->uses_dma;
 
-	rc = snprintf(name, NAMESZ, "%dt", ubus->busnum);
-	if (rc <= 0 || rc >= NAMESZ)
-		goto err_print_t;
-	d = debugfs_create_file(name, 0600, mondir, mbus, &mon_fops_text);
-	if (d == NULL)
-		goto err_create_t;
-	mbus->dent_t = d;
-
-	rc = snprintf(name, NAMESZ, "%ds", ubus->busnum);
-	if (rc <= 0 || rc >= NAMESZ)
-		goto err_print_s;
-	d = debugfs_create_file(name, 0600, mondir, mbus, &mon_fops_stat);
-	if (d == NULL)
-		goto err_create_s;
-	mbus->dent_s = d;
+	mbus->text_inited = mon_text_add(mbus, ubus);
+	// mon_bin_add(...)
 
 	mutex_lock(&mon_lock);
 	list_add_tail(&mbus->bus_link, &mon_buses);
 	mutex_unlock(&mon_lock);
 	return;
 
-err_create_s:
-err_print_s:
-	debugfs_remove(mbus->dent_t);
-err_create_t:
-err_print_t:
-	kfree(mbus);
 err_alloc:
 	return;
 }
 
+/*
+ * Search a USB bus by number. Notice that USB bus numbers start from one,
+ * which we may later use to identify "all" with zero.
+ *
+ * This function must be called with mon_lock held.
+ *
+ * This is obviously inefficient and may be revised in the future.
+ */
+struct mon_bus *mon_bus_lookup(unsigned int num)
+{
+	struct list_head *p;
+	struct mon_bus *mbus;
+
+	list_for_each (p, &mon_buses) {
+		mbus = list_entry(p, struct mon_bus, bus_link);
+		if (mbus->u_bus->busnum == num) {
+			return mbus;
+		}
+	}
+	return NULL;
+}
+
 static int __init mon_init(void)
 {
 	struct usb_bus *ubus;
-	struct dentry *mondir;
+	int rc;
 
-	mondir = debugfs_create_dir("usbmon", NULL);
-	if (IS_ERR(mondir)) {
-		printk(KERN_NOTICE TAG ": debugfs is not available\n");
-		return -ENODEV;
-	}
-	if (mondir == NULL) {
-		printk(KERN_NOTICE TAG ": unable to create usbmon directory\n");
-		return -ENODEV;
-	}
-	mon_dir = mondir;
+	if ((rc = mon_text_init()) != 0)
+		goto err_text;
+	if ((rc = mon_bin_init()) != 0)
+		goto err_bin;
 
 	if (usb_mon_register(&mon_ops_0) != 0) {
 		printk(KERN_NOTICE TAG ": unable to register with the core\n");
-		debugfs_remove(mondir);
-		return -ENODEV;
+		rc = -ENODEV;
+		goto err_reg;
 	}
 	// MOD_INC_USE_COUNT(which_module?);
 
@@ -361,10 +352,17 @@ static int __init mon_init(void)
 
 	mutex_lock(&usb_bus_list_lock);
 	list_for_each_entry (ubus, &usb_bus_list, bus_list) {
-		mon_bus_init(mondir, ubus);
+		mon_bus_init(ubus);
 	}
 	mutex_unlock(&usb_bus_list_lock);
 	return 0;
+
+err_reg:
+	mon_bin_exit();
+err_bin:
+	mon_text_exit();
+err_text:
+	return rc;
 }
 
 static void __exit mon_exit(void)
@@ -381,8 +379,8 @@ static void __exit mon_exit(void)
 		mbus = list_entry(p, struct mon_bus, bus_link);
 		list_del(p);
 
-		debugfs_remove(mbus->dent_t);
-		debugfs_remove(mbus->dent_s);
+		if (mbus->text_inited)
+			mon_text_del(mbus);
 
 		/*
 		 * This never happens, because the open/close paths in
@@ -401,7 +399,8 @@ static void __exit mon_exit(void)
 	}
 	mutex_unlock(&mon_lock);
 
-	debugfs_remove(mon_dir);
+	mon_text_exit();
+	mon_bin_exit();
 }
 
 module_init(mon_init);

drivers/usb/mon/mon_text.c

@@ -9,6 +9,7 @@
 #include <linux/usb.h>
 #include <linux/time.h>
 #include <linux/mutex.h>
+#include <linux/debugfs.h>
 #include <asm/uaccess.h>
 
 #include "usb_mon.h"
@@ -63,6 +64,8 @@ struct mon_reader_text {
 	char slab_name[SLAB_NAME_SZ];
 };
 
+static struct dentry *mon_dir;		/* Usually /sys/kernel/debug/usbmon */
+
 static void mon_text_ctor(void *, struct kmem_cache *, unsigned long);
 
 /*
@@ -436,7 +439,7 @@ static int mon_text_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
-const struct file_operations mon_fops_text = {
+static const struct file_operations mon_fops_text = {
 	.owner =	THIS_MODULE,
 	.open =		mon_text_open,
 	.llseek =	no_llseek,
@@ -447,6 +450,47 @@ const struct file_operations mon_fops_text = {
 	.release =	mon_text_release,
 };
 
+int mon_text_add(struct mon_bus *mbus, const struct usb_bus *ubus)
+{
+	struct dentry *d;
+	enum { NAMESZ = 10 };
+	char name[NAMESZ];
+	int rc;
+
+	rc = snprintf(name, NAMESZ, "%dt", ubus->busnum);
+	if (rc <= 0 || rc >= NAMESZ)
+		goto err_print_t;
+	d = debugfs_create_file(name, 0600, mon_dir, mbus, &mon_fops_text);
+	if (d == NULL)
+		goto err_create_t;
+	mbus->dent_t = d;
+
+	/* XXX The stats do not belong to here (text API), but oh well... */
+	rc = snprintf(name, NAMESZ, "%ds", ubus->busnum);
+	if (rc <= 0 || rc >= NAMESZ)
+		goto err_print_s;
+	d = debugfs_create_file(name, 0600, mon_dir, mbus, &mon_fops_stat);
+	if (d == NULL)
+		goto err_create_s;
+	mbus->dent_s = d;
+
+	return 1;
+
+err_create_s:
+err_print_s:
+	debugfs_remove(mbus->dent_t);
+	mbus->dent_t = NULL;
+err_create_t:
+err_print_t:
+	return 0;
+}
+
+void mon_text_del(struct mon_bus *mbus)
+{
+	debugfs_remove(mbus->dent_t);
+	debugfs_remove(mbus->dent_s);
+}
+
 /*
  * Slab interface: constructor.
  */
@@ -459,3 +503,24 @@ static void mon_text_ctor(void *mem, struct kmem_cache *slab, unsigned long sfla
 	memset(mem, 0xe5, sizeof(struct mon_event_text));
 }
 
+int __init mon_text_init(void)
+{
+	struct dentry *mondir;
+
+	mondir = debugfs_create_dir("usbmon", NULL);
+	if (IS_ERR(mondir)) {
+		printk(KERN_NOTICE TAG ": debugfs is not available\n");
+		return -ENODEV;
+	}
+	if (mondir == NULL) {
+		printk(KERN_NOTICE TAG ": unable to create usbmon directory\n");
+		return -ENODEV;
+	}
+	mon_dir = mondir;
+	return 0;
+}
+
+void __exit mon_text_exit(void)
+{
+	debugfs_remove(mon_dir);
+}

drivers/usb/mon/usb_mon.h

@@ -17,9 +17,11 @@
 struct mon_bus {
 	struct list_head bus_link;
 	spinlock_t lock;
+	struct usb_bus *u_bus;
+
+	int text_inited;
 	struct dentry *dent_s;		/* Debugging file */
 	struct dentry *dent_t;		/* Text interface file */
-	struct usb_bus *u_bus;
 	int uses_dma;
 
 	/* Ref */
@@ -48,13 +50,35 @@ struct mon_reader {
 void mon_reader_add(struct mon_bus *mbus, struct mon_reader *r);
 void mon_reader_del(struct mon_bus *mbus, struct mon_reader *r);
 
+struct mon_bus *mon_bus_lookup(unsigned int num);
+
+int /*bool*/ mon_text_add(struct mon_bus *mbus, const struct usb_bus *ubus);
+void mon_text_del(struct mon_bus *mbus);
+// void mon_bin_add(struct mon_bus *);
+
+int __init mon_text_init(void);
+void __exit mon_text_exit(void);
+int __init mon_bin_init(void);
+void __exit mon_bin_exit(void);
+
 /*
- */
+ * DMA interface.
+ *
+ * XXX The vectored side needs a serious re-thinking. Abstracting vectors,
+ * like in Paolo's original patch, produces a double pkmap. We need an idea.
+*/
 extern char mon_dmapeek(unsigned char *dst, dma_addr_t dma_addr, int len);
 
+struct mon_reader_bin;
+extern void mon_dmapeek_vec(const struct mon_reader_bin *rp,
+    unsigned int offset, dma_addr_t dma_addr, unsigned int len);
+extern unsigned int mon_copy_to_buff(const struct mon_reader_bin *rp,
+    unsigned int offset, const unsigned char *from, unsigned int len);
+
+/*
+ */
 extern struct mutex mon_lock;
 
-extern const struct file_operations mon_fops_text;
 extern const struct file_operations mon_fops_stat;
 
 #endif /* __USB_MON_H */