On Wed, 2011-03-09 at 00:42 +0000, Daniel Stodden wrote:
> Moves blktap2 definitions into a common header file.
>
> Includes xen/interface/io/ring.h and new ring definitions. Makes
> blktap build independently from xen-devel headers.
>
> New blktap_ring structs are fully congrent to blkif rings, for binary
> compat.
>
> Signed-off-by: Daniel Stodden <daniel.stodden@xxxxxxxxxx>
> ---
> drivers/xen/blktap/blktap.h | 66 ++++----------------------------
> drivers/xen/blktap/control.c | 14 +++---
> drivers/xen/blktap/device.c | 12 +++---
> drivers/xen/blktap/request.c | 8 ++--
> drivers/xen/blktap/ring.c | 51 ++++++++++++++-----------
> drivers/xen/blktap/sysfs.c | 6 +-
> include/linux/blktap.h | 85
> ++++++++++++++++++++++++++++++++++++++++++
This new file defines the kernel<->user (tapdisk process) ring protocol,
right?
I think its proper home would be under include/xen somewhere, which is
where the gntdev and evtchn etc driver interfaces are defined.
Where is the canonical definition of this interface stored? In the
kernel tree or the hypervisor tree?
Ian.
> 7 files changed, 142 insertions(+), 100 deletions(-)
> create mode 100644 include/linux/blktap.h
>
> diff --git a/drivers/xen/blktap/blktap.h b/drivers/xen/blktap/blktap.h
> index fe63fc9..1318cad 100644
> --- a/drivers/xen/blktap/blktap.h
> +++ b/drivers/xen/blktap/blktap.h
> @@ -6,7 +6,7 @@
> #include <linux/cdev.h>
> #include <linux/init.h>
> #include <linux/scatterlist.h>
> -#include <xen/blkif.h>
> +#include <linux/blktap.h>
>
> extern int blktap_debug_level;
> extern int blktap_ring_major;
> @@ -30,74 +30,26 @@ extern int blktap_device_major;
> #define BLKTAP_DEVICE_CLOSED 5
> #define BLKTAP_SHUTDOWN_REQUESTED 8
>
> -/* blktap IOCTLs: */
> -#define BLKTAP2_IOCTL_KICK_FE 1
> -#define BLKTAP2_IOCTL_ALLOC_TAP 200
> -#define BLKTAP2_IOCTL_FREE_TAP 201
> -#define BLKTAP2_IOCTL_CREATE_DEVICE 202
> -#define BLKTAP2_IOCTL_REMOVE_DEVICE 207
> -
> -#define BLKTAP2_MAX_MESSAGE_LEN 256
> -
> -#define BLKTAP2_RING_MESSAGE_CLOSE 3
> -
> #define BLKTAP_REQUEST_FREE 0
> #define BLKTAP_REQUEST_PENDING 1
>
> -/*
> - * The maximum number of requests that can be outstanding at any time
> - * is determined by
> - *
> - * [mmap_alloc * MAX_PENDING_REQS * BLKIF_MAX_SEGMENTS_PER_REQUEST]
> - *
> - * where mmap_alloc < MAX_DYNAMIC_MEM.
> - *
> - * TODO:
> - * mmap_alloc is initialised to 2 and should be adjustable on the fly via
> - * sysfs.
> - */
> -#define BLK_RING_SIZE __RING_SIZE((struct blkif_sring *)0,
> PAGE_SIZE)
> -#define MAX_DYNAMIC_MEM BLK_RING_SIZE
> -#define MAX_PENDING_REQS BLK_RING_SIZE
> -#define MMAP_PAGES (MAX_PENDING_REQS * BLKIF_MAX_SEGMENTS_PER_REQUEST)
> -#define MMAP_VADDR(_start, _req, _seg) \
> - (_start + \
> - ((_req) * BLKIF_MAX_SEGMENTS_PER_REQUEST * PAGE_SIZE) + \
> - ((_seg) * PAGE_SIZE))
> -
> -struct grant_handle_pair {
> - grant_handle_t kernel;
> - grant_handle_t user;
> -};
> -#define INVALID_GRANT_HANDLE 0xFFFF
> -
> -struct blktap_handle {
> - unsigned int ring;
> - unsigned int device;
> - unsigned int minor;
> -};
> -
> -struct blktap_params {
> - char name[BLKTAP2_MAX_MESSAGE_LEN];
> - unsigned long long capacity;
> - unsigned long sector_size;
> -};
> -
> struct blktap_device {
> spinlock_t lock;
> struct gendisk *gd;
> };
>
> +struct blktap_request;
> +
> struct blktap_ring {
> struct task_struct *task;
>
> struct vm_area_struct *vma;
> - struct blkif_front_ring ring;
> + blktap_front_ring_t ring;
> unsigned long ring_vstart;
> unsigned long user_vstart;
>
> int n_pending;
> - struct blktap_request *pending[MAX_PENDING_REQS];
> + struct blktap_request *pending[BLKTAP_RING_SIZE];
>
> wait_queue_head_t poll_wait;
>
> @@ -128,8 +80,8 @@ struct blktap_request {
> int operation;
> struct timeval time;
>
> - struct scatterlist
> sg_table[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> - struct page *pages[BLKIF_MAX_SEGMENTS_PER_REQUEST];
> + struct scatterlist sg_table[BLKTAP_SEGMENT_MAX];
> + struct page *pages[BLKTAP_SEGMENT_MAX];
> int nr_pages;
> };
>
> @@ -148,7 +100,7 @@ struct blktap {
>
> wait_queue_head_t remove_wait;
> struct work_struct remove_work;
> - char name[BLKTAP2_MAX_MESSAGE_LEN];
> + char name[BLKTAP_NAME_MAX];
>
> struct blktap_statistics stats;
> };
> @@ -189,7 +141,7 @@ void blktap_sysfs_destroy(struct blktap *);
> int blktap_device_init(void);
> void blktap_device_exit(void);
> size_t blktap_device_debug(struct blktap *, char *, size_t);
> -int blktap_device_create(struct blktap *, struct blktap_params *);
> +int blktap_device_create(struct blktap *, struct blktap_device_info *);
> int blktap_device_destroy(struct blktap *);
> void blktap_device_destroy_sync(struct blktap *);
> void blktap_device_run_queue(struct blktap *);
> diff --git a/drivers/xen/blktap/control.c b/drivers/xen/blktap/control.c
> index f339bba..57b1a10 100644
> --- a/drivers/xen/blktap/control.c
> +++ b/drivers/xen/blktap/control.c
> @@ -127,19 +127,19 @@ blktap_control_ioctl(struct inode *inode, struct file
> *filp,
> struct blktap *tap;
>
> switch (cmd) {
> - case BLKTAP2_IOCTL_ALLOC_TAP: {
> - struct blktap_handle h;
> + case BLKTAP_IOCTL_ALLOC_TAP: {
> + struct blktap_info info;
> void __user *ptr = (void __user*)arg;
>
> tap = blktap_control_create_tap();
> if (!tap)
> return -ENOMEM;
>
> - h.ring = blktap_ring_major;
> - h.device = blktap_device_major;
> - h.minor = tap->minor;
> + info.ring_major = blktap_ring_major;
> + info.bdev_major = blktap_device_major;
> + info.ring_minor = tap->minor;
>
> - if (copy_to_user(ptr, &h, sizeof(h))) {
> + if (copy_to_user(ptr, &info, sizeof(info))) {
> blktap_control_destroy_tap(tap);
> return -EFAULT;
> }
> @@ -147,7 +147,7 @@ blktap_control_ioctl(struct inode *inode, struct file
> *filp,
> return 0;
> }
>
> - case BLKTAP2_IOCTL_FREE_TAP: {
> + case BLKTAP_IOCTL_FREE_TAP: {
> int minor = arg;
>
> if (minor > MAX_BLKTAP_DEVICE)
> diff --git a/drivers/xen/blktap/device.c b/drivers/xen/blktap/device.c
> index fce2769..6bb04bd 100644
> --- a/drivers/xen/blktap/device.c
> +++ b/drivers/xen/blktap/device.c
> @@ -186,7 +186,7 @@ blktap_device_make_request(struct blktap *tap, struct
> request *rq)
> write ? 'w' : 'r', blk_rq_bytes(rq), nsegs);
>
> request->rq = rq;
> - request->operation = write ? BLKIF_OP_WRITE : BLKIF_OP_READ;
> + request->operation = write ? BLKTAP_OP_WRITE : BLKTAP_OP_READ;
>
> err = blktap_request_get_pages(tap, request, nsegs);
> if (err)
> @@ -276,7 +276,7 @@ blktap_device_do_request(struct request_queue *rq)
>
> static void
> blktap_device_configure(struct blktap *tap,
> - struct blktap_params *params)
> + struct blktap_device_info *params)
> {
> struct request_queue *rq;
> struct blktap_device *dev = &tap->device;
> @@ -297,8 +297,8 @@ blktap_device_configure(struct blktap *tap,
> blk_queue_max_segment_size(rq, PAGE_SIZE);
>
> /* Ensure a merged request will fit in a single I/O ring slot. */
> - blk_queue_max_phys_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
> - blk_queue_max_hw_segments(rq, BLKIF_MAX_SEGMENTS_PER_REQUEST);
> + blk_queue_max_phys_segments(rq, BLKTAP_SEGMENT_MAX);
> + blk_queue_max_hw_segments(rq, BLKTAP_SEGMENT_MAX);
>
> /* Make sure buffer addresses are sector-aligned. */
> blk_queue_dma_alignment(rq, 511);
> @@ -311,7 +311,7 @@ blktap_device_configure(struct blktap *tap,
>
> static int
> blktap_device_validate_params(struct blktap *tap,
> - struct blktap_params *params)
> + struct blktap_device_info *params)
> {
> struct device *dev = tap->ring.dev;
> int sector_order, name_sz;
> @@ -425,7 +425,7 @@ blktap_device_destroy_sync(struct blktap *tap)
> }
>
> int
> -blktap_device_create(struct blktap *tap, struct blktap_params *params)
> +blktap_device_create(struct blktap *tap, struct blktap_device_info *params)
> {
> int minor, err;
> struct gendisk *gd;
> diff --git a/drivers/xen/blktap/request.c b/drivers/xen/blktap/request.c
> index 9bef48c..8cfd6c9 100644
> --- a/drivers/xen/blktap/request.c
> +++ b/drivers/xen/blktap/request.c
> @@ -7,18 +7,18 @@
> #include "blktap.h"
>
> /* max pages per shared pool. just to prevent accidental dos. */
> -#define POOL_MAX_PAGES (256*BLKIF_MAX_SEGMENTS_PER_REQUEST)
> +#define POOL_MAX_PAGES (256*BLKTAP_SEGMENT_MAX)
>
> /* default page pool size. when considering to shrink a shared pool,
> * note that paused tapdisks may grab a whole lot of pages for a long
> * time. */
> -#define POOL_DEFAULT_PAGES (2 * MMAP_PAGES)
> +#define POOL_DEFAULT_PAGES (2 * BLKTAP_RING_SIZE * BLKTAP_SEGMENT_MAX)
>
> /* max number of pages allocatable per request. */
> -#define POOL_MAX_REQUEST_PAGES BLKIF_MAX_SEGMENTS_PER_REQUEST
> +#define POOL_MAX_REQUEST_PAGES BLKTAP_SEGMENT_MAX
>
> /* min request structs per pool. These grow dynamically. */
> -#define POOL_MIN_REQS BLK_RING_SIZE
> +#define POOL_MIN_REQS BLKTAP_RING_SIZE
>
> static struct kset *pool_set;
>
> diff --git a/drivers/xen/blktap/ring.c b/drivers/xen/blktap/ring.c
> index 6b86be5..9442a64 100644
> --- a/drivers/xen/blktap/ring.c
> +++ b/drivers/xen/blktap/ring.c
> @@ -18,7 +18,7 @@ static struct cdev blktap_ring_cdev;
>
> static void
> blktap_ring_read_response(struct blktap *tap,
> - const struct blkif_response *rsp)
> + const blktap_ring_rsp_t *rsp)
> {
> struct blktap_ring *ring = &tap->ring;
> struct blktap_request *request;
> @@ -27,7 +27,7 @@ blktap_ring_read_response(struct blktap *tap,
> request = NULL;
>
> usr_idx = rsp->id;
> - if (usr_idx < 0 || usr_idx >= MAX_PENDING_REQS) {
> + if (usr_idx < 0 || usr_idx >= BLKTAP_RING_SIZE) {
> err = -ERANGE;
> goto invalid;
> }
> @@ -48,7 +48,7 @@ blktap_ring_read_response(struct blktap *tap,
> "request %d [%p] response: %d\n",
> request->usr_idx, request, rsp->status);
>
> - err = rsp->status == BLKIF_RSP_OKAY ? 0 : -EIO;
> + err = rsp->status == BLKTAP_RSP_OKAY ? 0 : -EIO;
> end_request:
> blktap_device_end_request(tap, request, err);
> return;
> @@ -67,7 +67,7 @@ static void
> blktap_read_ring(struct blktap *tap)
> {
> struct blktap_ring *ring = &tap->ring;
> - struct blkif_response rsp;
> + blktap_ring_rsp_t rsp;
> RING_IDX rc, rp;
>
> down_read(¤t->mm->mmap_sem);
> @@ -90,6 +90,11 @@ blktap_read_ring(struct blktap *tap)
> up_read(¤t->mm->mmap_sem);
> }
>
> +#define MMAP_VADDR(_start, _req, _seg) \
> + ((_start) + \
> + ((_req) * BLKTAP_SEGMENT_MAX * BLKTAP_PAGE_SIZE) + \
> + ((_seg) * BLKTAP_PAGE_SIZE))
> +
> static int blktap_ring_fault(struct vm_area_struct *vma, struct vm_fault
> *vmf)
> {
> return VM_FAULT_SIGBUS;
> @@ -102,7 +107,7 @@ blktap_ring_fail_pending(struct blktap *tap)
> struct blktap_request *request;
> int usr_idx;
>
> - for (usr_idx = 0; usr_idx < MAX_PENDING_REQS; usr_idx++) {
> + for (usr_idx = 0; usr_idx < BLKTAP_RING_SIZE; usr_idx++) {
> request = ring->pending[usr_idx];
> if (!request)
> continue;
> @@ -154,7 +159,7 @@ blktap_ring_map_request(struct blktap *tap,
> int seg, err = 0;
> int write;
>
> - write = request->operation == BLKIF_OP_WRITE;
> + write = request->operation == BLKTAP_OP_WRITE;
>
> for (seg = 0; seg < request->nr_pages; seg++) {
> if (write)
> @@ -182,7 +187,7 @@ blktap_ring_unmap_request(struct blktap *tap,
>
> uaddr = MMAP_VADDR(ring->user_vstart, request->usr_idx, 0);
> size = request->nr_pages << PAGE_SHIFT;
> - read = request->operation == BLKIF_OP_READ;
> + read = request->operation == BLKTAP_OP_READ;
>
> if (read)
> for (seg = 0; seg < request->nr_pages; seg++)
> @@ -217,11 +222,11 @@ blktap_ring_make_request(struct blktap *tap)
> if (!request)
> return ERR_PTR(-ENOMEM);
>
> - for (usr_idx = 0; usr_idx < BLK_RING_SIZE; usr_idx++)
> + for (usr_idx = 0; usr_idx < BLKTAP_RING_SIZE; usr_idx++)
> if (!ring->pending[usr_idx])
> break;
>
> - BUG_ON(usr_idx >= BLK_RING_SIZE);
> + BUG_ON(usr_idx >= BLKTAP_RING_SIZE);
>
> request->tap = tap;
> request->usr_idx = usr_idx;
> @@ -237,7 +242,7 @@ blktap_ring_submit_request(struct blktap *tap,
> struct blktap_request *request)
> {
> struct blktap_ring *ring = &tap->ring;
> - struct blkif_request *breq;
> + blktap_ring_req_t *breq;
> struct scatterlist *sg;
> int i, nsecs = 0;
>
> @@ -248,12 +253,12 @@ blktap_ring_submit_request(struct blktap *tap,
>
> breq->id = request->usr_idx;
> breq->sector_number = blk_rq_pos(request->rq);
> - breq->handle = 0;
> + breq->__pad = 0;
> breq->operation = request->operation;
> breq->nr_segments = request->nr_pages;
>
> blktap_for_each_sg(sg, request, i) {
> - struct blkif_request_segment *seg = &breq->seg[i];
> + struct blktap_segment *seg = &breq->seg[i];
> int first, count;
>
> count = sg->length >> 9;
> @@ -270,12 +275,12 @@ blktap_ring_submit_request(struct blktap *tap,
> do_gettimeofday(&request->time);
>
>
> - if (request->operation == BLKIF_OP_WRITE) {
> + if (request->operation == BLKTAP_OP_WRITE) {
> tap->stats.st_wr_sect += nsecs;
> tap->stats.st_wr_req++;
> }
>
> - if (request->operation == BLKIF_OP_READ) {
> + if (request->operation == BLKTAP_OP_READ) {
> tap->stats.st_rd_sect += nsecs;
> tap->stats.st_rd_req++;
> }
> @@ -327,7 +332,7 @@ blktap_ring_mmap(struct file *filp, struct vm_area_struct
> *vma)
> {
> struct blktap *tap = filp->private_data;
> struct blktap_ring *ring = &tap->ring;
> - struct blkif_sring *sring;
> + blktap_sring_t *sring;
> struct page *page = NULL;
> int err;
>
> @@ -384,25 +389,25 @@ blktap_ring_ioctl(struct inode *inode, struct file
> *filp,
> return -EACCES;
>
> switch(cmd) {
> - case BLKTAP2_IOCTL_KICK_FE:
> + case BLKTAP_IOCTL_RESPOND:
>
> blktap_read_ring(tap);
> return 0;
>
> - case BLKTAP2_IOCTL_CREATE_DEVICE: {
> - struct blktap_params params;
> + case BLKTAP_IOCTL_CREATE_DEVICE: {
> + struct blktap_device_info info;
> void __user *ptr = (void *)arg;
>
> if (!arg)
> return -EINVAL;
>
> - if (copy_from_user(¶ms, ptr, sizeof(params)))
> + if (copy_from_user(&info, ptr, sizeof(info)))
> return -EFAULT;
>
> - return blktap_device_create(tap, ¶ms);
> + return blktap_device_create(tap, &info);
> }
>
> - case BLKTAP2_IOCTL_REMOVE_DEVICE:
> + case BLKTAP_IOCTL_REMOVE_DEVICE:
>
> return blktap_device_destroy(tap);
> }
> @@ -482,7 +487,7 @@ blktap_ring_debug(struct blktap *tap, char *buf, size_t
> size)
> s += snprintf(s, end - s,
> "begin pending:%d\n", ring->n_pending);
>
> - for (usr_idx = 0; usr_idx < MAX_PENDING_REQS; usr_idx++) {
> + for (usr_idx = 0; usr_idx < BLKTAP_RING_SIZE; usr_idx++) {
> struct blktap_request *request;
> struct timeval *time;
> int write;
> @@ -491,7 +496,7 @@ blktap_ring_debug(struct blktap *tap, char *buf, size_t
> size)
> if (!request)
> continue;
>
> - write = request->operation == BLKIF_OP_WRITE;
> + write = request->operation == BLKTAP_OP_WRITE;
> time = &request->time;
>
> s += snprintf(s, end - s,
> diff --git a/drivers/xen/blktap/sysfs.c b/drivers/xen/blktap/sysfs.c
> index 7bbfea8..182de9a 100644
> --- a/drivers/xen/blktap/sysfs.c
> +++ b/drivers/xen/blktap/sysfs.c
> @@ -20,7 +20,7 @@ blktap_sysfs_set_name(struct device *dev, struct
> device_attribute *attr, const c
> if (!tap)
> return 0;
>
> - if (size >= BLKTAP2_MAX_MESSAGE_LEN)
> + if (size >= BLKTAP_NAME_MAX)
> return -ENAMETOOLONG;
>
> if (strnlen(buf, size) != size)
> @@ -75,8 +75,8 @@ blktap_sysfs_remove_device(struct device *dev,
> goto wait;
>
> if (tap->ring.vma) {
> - blkif_sring_t *sring = tap->ring.ring.sring;
> - sring->private.tapif_user.msg = BLKTAP2_RING_MESSAGE_CLOSE;
> + blktap_sring_t *sring = tap->ring.ring.sring;
> + sring->private.tapif_user.msg = BLKTAP_RING_MESSAGE_CLOSE;
> blktap_ring_kick_user(tap);
> } else {
> INIT_WORK(&tap->remove_work, blktap_sysfs_remove_work);
> diff --git a/include/linux/blktap.h b/include/linux/blktap.h
> new file mode 100644
> index 0000000..ec33429
> --- /dev/null
> +++ b/include/linux/blktap.h
> @@ -0,0 +1,85 @@
> +/*
> + * Copyright (c) 2011, XenSource Inc.
> + * All rights reserved.
> + */
> +
> +#ifndef _LINUX_BLKTAP_H
> +#define _LINUX_BLKTAP_H
> +
> +/*
> + * Control
> + */
> +
> +#define BLKTAP_IOCTL_RESPOND 1
> +#define BLKTAP_IOCTL_ALLOC_TAP 200
> +#define BLKTAP_IOCTL_FREE_TAP 201
> +#define BLKTAP_IOCTL_CREATE_DEVICE 202
> +#define BLKTAP_IOCTL_REMOVE_DEVICE 207
> +
> +#define BLKTAP_NAME_MAX 256
> +
> +struct blktap_info {
> + unsigned int ring_major;
> + unsigned int bdev_major;
> + unsigned int ring_minor;
> +};
> +
> +struct blktap_device_info {
> + char name[BLKTAP_NAME_MAX];
> + unsigned long long capacity;
> + unsigned long sector_size;
> +};
> +
> +/*
> + * I/O ring
> + */
> +
> +#ifdef __KERNEL__
> +#define BLKTAP_PAGE_SIZE PAGE_SIZE
> +#endif
> +
> +#include <xen/interface/io/ring.h>
> +
> +typedef struct blktap_ring_request blktap_ring_req_t;
> +typedef struct blktap_ring_response blktap_ring_rsp_t;
> +
> +struct blktap_segment {
> + uint32_t __pad;
> + uint8_t first_sect;
> + uint8_t last_sect;
> +};
> +
> +#define BLKTAP_OP_READ 0
> +#define BLKTAP_OP_WRITE 1
> +
> +#define BLKTAP_SEGMENT_MAX 11
> +
> +struct blktap_ring_request {
> + uint8_t operation;
> + uint8_t nr_segments;
> + uint16_t __pad;
> + uint64_t id;
> + uint64_t sector_number;
> + struct blktap_segment seg[BLKTAP_SEGMENT_MAX];
> +};
> +
> +#define BLKTAP_RSP_EOPNOTSUPP -2
> +#define BLKTAP_RSP_ERROR -1
> +#define BLKTAP_RSP_OKAY 0
> +
> +struct blktap_ring_response {
> + uint64_t id;
> + uint8_t operation;
> + int16_t status;
> +};
> +
> +DEFINE_RING_TYPES(blktap, struct blktap_ring_request, struct
> blktap_ring_response);
> +#define BLKTAP_RING_SIZE __CONST_RING_SIZE(blktap, BLKTAP_PAGE_SIZE)
> +
> +/*
> + * Ring messages (DEPRECATED)
> + */
> +
> +#define BLKTAP_RING_MESSAGE_CLOSE 3
> +
> +#endif /* _LINUX_BLKTAP_H */
> --
> 1.7.0.4
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home