2011-04-15 05:25:47 +07:00
|
|
|
/*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License version 2
|
|
|
|
* as published by the Free Software Foundation; or, when distributed
|
|
|
|
* separately from the Linux kernel or incorporated into other
|
|
|
|
* software packages, subject to the following license:
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
|
|
* of this source file (the "Software"), to deal in the Software without
|
|
|
|
* restriction, including without limitation the rights to use, copy, modify,
|
|
|
|
* merge, publish, distribute, sublicense, and/or sell copies of the Software,
|
|
|
|
* and to permit persons to whom the Software is furnished to do so, subject to
|
|
|
|
* the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice shall be included in
|
|
|
|
* all copies or substantial portions of the Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
|
|
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
|
|
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
|
|
|
|
* IN THE SOFTWARE.
|
|
|
|
*/
|
|
|
|
|
2011-05-13 03:58:21 +07:00
|
|
|
#ifndef __XEN_BLKIF__BACKEND__COMMON_H__
|
|
|
|
#define __XEN_BLKIF__BACKEND__COMMON_H__
|
2011-04-15 05:25:47 +07:00
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/vmalloc.h>
|
|
|
|
#include <linux/wait.h>
|
2011-04-15 04:21:50 +07:00
|
|
|
#include <linux/io.h>
|
xen/blkback: Persistent grant maps for xen blk drivers
This patch implements persistent grants for the xen-blk{front,back}
mechanism. The effect of this change is to reduce the number of unmap
operations performed, since they cause a (costly) TLB shootdown. This
allows the I/O performance to scale better when a large number of VMs
are performing I/O.
Previously, the blkfront driver was supplied a bvec[] from the request
queue. This was granted to dom0; dom0 performed the I/O and wrote
directly into the grant-mapped memory and unmapped it; blkfront then
removed foreign access for that grant. The cost of unmapping scales
badly with the number of CPUs in Dom0. An experiment showed that when
Dom0 has 24 VCPUs, and guests are performing parallel I/O to a
ramdisk, the IPIs from performing unmap's is a bottleneck at 5 guests
(at which point 650,000 IOPS are being performed in total). If more
than 5 guests are used, the performance declines. By 10 guests, only
400,000 IOPS are being performed.
This patch improves performance by only unmapping when the connection
between blkfront and back is broken.
On startup blkfront notifies blkback that it is using persistent
grants, and blkback will do the same. If blkback is not capable of
persistent mapping, blkfront will still use the same grants, since it
is compatible with the previous protocol, and simplifies the code
complexity in blkfront.
To perform a read, in persistent mode, blkfront uses a separate pool
of pages that it maps to dom0. When a request comes in, blkfront
transmutes the request so that blkback will write into one of these
free pages. Blkback keeps note of which grefs it has already
mapped. When a new ring request comes to blkback, it looks to see if
it has already mapped that page. If so, it will not map it again. If
the page hasn't been previously mapped, it is mapped now, and a record
is kept of this mapping. Blkback proceeds as usual. When blkfront is
notified that blkback has completed a request, it memcpy's from the
shared memory, into the bvec supplied. A record that the {gref, page}
tuple is mapped, and not inflight is kept.
Writes are similar, except that the memcpy is peformed from the
supplied bvecs, into the shared pages, before the request is put onto
the ring.
Blkback stores a mapping of grefs=>{page mapped to by gref} in
a red-black tree. As the grefs are not known apriori, and provide no
guarantees on their ordering, we have to perform a search
through this tree to find the page, for every gref we receive. This
operation takes O(log n) time in the worst case. In blkfront grants
are stored using a single linked list.
The maximum number of grants that blkback will persistenly map is
currently set to RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, to
prevent a malicios guest from attempting a DoS, by supplying fresh
grefs, causing the Dom0 kernel to map excessively. If a guest
is using persistent grants and exceeds the maximum number of grants to
map persistenly the newly passed grefs will be mapped and unmaped.
Using this approach, we can have requests that mix persistent and
non-persistent grants, and we need to handle them correctly.
This allows us to set the maximum number of persistent grants to a
lower value than RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, although
setting it will lead to unpredictable performance.
In writing this patch, the question arrises as to if the additional
cost of performing memcpys in the guest (to/from the pool of granted
pages) outweigh the gains of not performing TLB shootdowns. The answer
to that question is `no'. There appears to be very little, if any
additional cost to the guest of using persistent grants. There is
perhaps a small saving, from the reduced number of hypercalls
performed in granting, and ending foreign access.
Signed-off-by: Oliver Chick <oliver.chick@citrix.com>
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[v1: Fixed up the misuse of bool as int]
2012-10-24 23:58:45 +07:00
|
|
|
#include <linux/rbtree.h>
|
2011-04-15 05:25:47 +07:00
|
|
|
#include <asm/setup.h>
|
|
|
|
#include <asm/pgalloc.h>
|
|
|
|
#include <asm/hypervisor.h>
|
2009-02-10 03:05:51 +07:00
|
|
|
#include <xen/grant_table.h>
|
2011-04-15 05:25:47 +07:00
|
|
|
#include <xen/xenbus.h>
|
2011-05-13 03:31:51 +07:00
|
|
|
#include <xen/interface/io/ring.h>
|
|
|
|
#include <xen/interface/io/blkif.h>
|
|
|
|
#include <xen/interface/io/protocols.h>
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-05-13 03:43:12 +07:00
|
|
|
#define DRV_PFX "xen-blkback:"
|
2011-05-12 03:15:24 +07:00
|
|
|
#define DPRINTK(fmt, args...) \
|
2011-08-15 11:57:07 +07:00
|
|
|
pr_debug(DRV_PFX "(%s:%d) " fmt ".\n", \
|
2011-05-12 03:15:24 +07:00
|
|
|
__func__, __LINE__, ##args)
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-05-13 03:31:51 +07:00
|
|
|
|
|
|
|
/* Not a real protocol. Used to generate ring structs which contain
|
|
|
|
* the elements common to all protocols only. This way we get a
|
|
|
|
* compiler-checkable way to use common struct elements, so we can
|
|
|
|
* avoid using switch(protocol) in a number of places. */
|
|
|
|
struct blkif_common_request {
|
|
|
|
char dummy;
|
|
|
|
};
|
|
|
|
struct blkif_common_response {
|
|
|
|
char dummy;
|
|
|
|
};
|
|
|
|
|
2011-09-01 17:39:10 +07:00
|
|
|
struct blkif_x86_32_request_rw {
|
2011-10-12 23:12:36 +07:00
|
|
|
uint8_t nr_segments; /* number of segments */
|
|
|
|
blkif_vdev_t handle; /* only for read/write requests */
|
|
|
|
uint64_t id; /* private guest value, echoed in resp */
|
2011-09-01 17:39:10 +07:00
|
|
|
blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
|
|
|
|
struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
|
2011-10-12 23:12:36 +07:00
|
|
|
} __attribute__((__packed__));
|
2011-09-01 17:39:10 +07:00
|
|
|
|
|
|
|
struct blkif_x86_32_request_discard {
|
2011-10-13 03:23:30 +07:00
|
|
|
uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */
|
2011-10-12 23:12:36 +07:00
|
|
|
blkif_vdev_t _pad1; /* was "handle" for read/write requests */
|
|
|
|
uint64_t id; /* private guest value, echoed in resp */
|
2011-09-01 17:39:10 +07:00
|
|
|
blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
|
2011-10-12 23:12:36 +07:00
|
|
|
uint64_t nr_sectors;
|
|
|
|
} __attribute__((__packed__));
|
2011-09-01 17:39:10 +07:00
|
|
|
|
2011-05-13 03:31:51 +07:00
|
|
|
struct blkif_x86_32_request {
|
|
|
|
uint8_t operation; /* BLKIF_OP_??? */
|
2011-09-01 17:39:10 +07:00
|
|
|
union {
|
|
|
|
struct blkif_x86_32_request_rw rw;
|
|
|
|
struct blkif_x86_32_request_discard discard;
|
|
|
|
} u;
|
2011-10-12 23:12:36 +07:00
|
|
|
} __attribute__((__packed__));
|
|
|
|
|
|
|
|
/* i386 protocol version */
|
|
|
|
#pragma pack(push, 4)
|
2011-05-13 03:31:51 +07:00
|
|
|
struct blkif_x86_32_response {
|
|
|
|
uint64_t id; /* copied from request */
|
|
|
|
uint8_t operation; /* copied from request */
|
|
|
|
int16_t status; /* BLKIF_RSP_??? */
|
|
|
|
};
|
|
|
|
#pragma pack(pop)
|
|
|
|
/* x86_64 protocol version */
|
2011-09-01 17:39:10 +07:00
|
|
|
|
|
|
|
struct blkif_x86_64_request_rw {
|
2011-10-12 23:12:36 +07:00
|
|
|
uint8_t nr_segments; /* number of segments */
|
|
|
|
blkif_vdev_t handle; /* only for read/write requests */
|
|
|
|
uint32_t _pad1; /* offsetof(blkif_reqest..,u.rw.id)==8 */
|
|
|
|
uint64_t id;
|
2011-09-01 17:39:10 +07:00
|
|
|
blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
|
|
|
|
struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
|
2011-10-12 23:12:36 +07:00
|
|
|
} __attribute__((__packed__));
|
2011-09-01 17:39:10 +07:00
|
|
|
|
|
|
|
struct blkif_x86_64_request_discard {
|
2011-10-13 03:23:30 +07:00
|
|
|
uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */
|
2011-10-12 23:12:36 +07:00
|
|
|
blkif_vdev_t _pad1; /* was "handle" for read/write requests */
|
|
|
|
uint32_t _pad2; /* offsetof(blkif_..,u.discard.id)==8 */
|
|
|
|
uint64_t id;
|
2011-09-01 17:39:10 +07:00
|
|
|
blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
|
2011-10-12 23:12:36 +07:00
|
|
|
uint64_t nr_sectors;
|
|
|
|
} __attribute__((__packed__));
|
2011-09-01 17:39:10 +07:00
|
|
|
|
2011-05-13 03:31:51 +07:00
|
|
|
struct blkif_x86_64_request {
|
|
|
|
uint8_t operation; /* BLKIF_OP_??? */
|
2011-09-01 17:39:10 +07:00
|
|
|
union {
|
|
|
|
struct blkif_x86_64_request_rw rw;
|
|
|
|
struct blkif_x86_64_request_discard discard;
|
|
|
|
} u;
|
2011-10-12 23:12:36 +07:00
|
|
|
} __attribute__((__packed__));
|
|
|
|
|
2011-05-13 03:31:51 +07:00
|
|
|
struct blkif_x86_64_response {
|
|
|
|
uint64_t __attribute__((__aligned__(8))) id;
|
|
|
|
uint8_t operation; /* copied from request */
|
|
|
|
int16_t status; /* BLKIF_RSP_??? */
|
|
|
|
};
|
|
|
|
|
|
|
|
DEFINE_RING_TYPES(blkif_common, struct blkif_common_request,
|
|
|
|
struct blkif_common_response);
|
|
|
|
DEFINE_RING_TYPES(blkif_x86_32, struct blkif_x86_32_request,
|
|
|
|
struct blkif_x86_32_response);
|
|
|
|
DEFINE_RING_TYPES(blkif_x86_64, struct blkif_x86_64_request,
|
|
|
|
struct blkif_x86_64_response);
|
|
|
|
|
|
|
|
union blkif_back_rings {
|
|
|
|
struct blkif_back_ring native;
|
|
|
|
struct blkif_common_back_ring common;
|
|
|
|
struct blkif_x86_32_back_ring x86_32;
|
|
|
|
struct blkif_x86_64_back_ring x86_64;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum blkif_protocol {
|
|
|
|
BLKIF_PROTOCOL_NATIVE = 1,
|
|
|
|
BLKIF_PROTOCOL_X86_32 = 2,
|
|
|
|
BLKIF_PROTOCOL_X86_64 = 3,
|
|
|
|
};
|
|
|
|
|
2011-05-13 03:53:56 +07:00
|
|
|
struct xen_vbd {
|
2011-05-12 02:57:09 +07:00
|
|
|
/* What the domain refers to this vbd as. */
|
|
|
|
blkif_vdev_t handle;
|
|
|
|
/* Non-zero -> read-only */
|
|
|
|
unsigned char readonly;
|
|
|
|
/* VDISK_xxx */
|
|
|
|
unsigned char type;
|
|
|
|
/* phys device that this vbd maps to. */
|
|
|
|
u32 pdevice;
|
|
|
|
struct block_device *bdev;
|
|
|
|
/* Cached size parameter. */
|
|
|
|
sector_t size;
|
2012-09-21 16:04:18 +07:00
|
|
|
unsigned int flush_support:1;
|
|
|
|
unsigned int discard_secure:1;
|
xen/blkback: Persistent grant maps for xen blk drivers
This patch implements persistent grants for the xen-blk{front,back}
mechanism. The effect of this change is to reduce the number of unmap
operations performed, since they cause a (costly) TLB shootdown. This
allows the I/O performance to scale better when a large number of VMs
are performing I/O.
Previously, the blkfront driver was supplied a bvec[] from the request
queue. This was granted to dom0; dom0 performed the I/O and wrote
directly into the grant-mapped memory and unmapped it; blkfront then
removed foreign access for that grant. The cost of unmapping scales
badly with the number of CPUs in Dom0. An experiment showed that when
Dom0 has 24 VCPUs, and guests are performing parallel I/O to a
ramdisk, the IPIs from performing unmap's is a bottleneck at 5 guests
(at which point 650,000 IOPS are being performed in total). If more
than 5 guests are used, the performance declines. By 10 guests, only
400,000 IOPS are being performed.
This patch improves performance by only unmapping when the connection
between blkfront and back is broken.
On startup blkfront notifies blkback that it is using persistent
grants, and blkback will do the same. If blkback is not capable of
persistent mapping, blkfront will still use the same grants, since it
is compatible with the previous protocol, and simplifies the code
complexity in blkfront.
To perform a read, in persistent mode, blkfront uses a separate pool
of pages that it maps to dom0. When a request comes in, blkfront
transmutes the request so that blkback will write into one of these
free pages. Blkback keeps note of which grefs it has already
mapped. When a new ring request comes to blkback, it looks to see if
it has already mapped that page. If so, it will not map it again. If
the page hasn't been previously mapped, it is mapped now, and a record
is kept of this mapping. Blkback proceeds as usual. When blkfront is
notified that blkback has completed a request, it memcpy's from the
shared memory, into the bvec supplied. A record that the {gref, page}
tuple is mapped, and not inflight is kept.
Writes are similar, except that the memcpy is peformed from the
supplied bvecs, into the shared pages, before the request is put onto
the ring.
Blkback stores a mapping of grefs=>{page mapped to by gref} in
a red-black tree. As the grefs are not known apriori, and provide no
guarantees on their ordering, we have to perform a search
through this tree to find the page, for every gref we receive. This
operation takes O(log n) time in the worst case. In blkfront grants
are stored using a single linked list.
The maximum number of grants that blkback will persistenly map is
currently set to RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, to
prevent a malicios guest from attempting a DoS, by supplying fresh
grefs, causing the Dom0 kernel to map excessively. If a guest
is using persistent grants and exceeds the maximum number of grants to
map persistenly the newly passed grefs will be mapped and unmaped.
Using this approach, we can have requests that mix persistent and
non-persistent grants, and we need to handle them correctly.
This allows us to set the maximum number of persistent grants to a
lower value than RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, although
setting it will lead to unpredictable performance.
In writing this patch, the question arrises as to if the additional
cost of performing memcpys in the guest (to/from the pool of granted
pages) outweigh the gains of not performing TLB shootdowns. The answer
to that question is `no'. There appears to be very little, if any
additional cost to the guest of using persistent grants. There is
perhaps a small saving, from the reduced number of hypercalls
performed in granting, and ending foreign access.
Signed-off-by: Oliver Chick <oliver.chick@citrix.com>
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[v1: Fixed up the misuse of bool as int]
2012-10-24 23:58:45 +07:00
|
|
|
unsigned int feature_gnt_persistent:1;
|
|
|
|
unsigned int overflow_max_grants:1;
|
2011-04-15 05:25:47 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct backend_info;
|
|
|
|
|
xen/blkback: Persistent grant maps for xen blk drivers
This patch implements persistent grants for the xen-blk{front,back}
mechanism. The effect of this change is to reduce the number of unmap
operations performed, since they cause a (costly) TLB shootdown. This
allows the I/O performance to scale better when a large number of VMs
are performing I/O.
Previously, the blkfront driver was supplied a bvec[] from the request
queue. This was granted to dom0; dom0 performed the I/O and wrote
directly into the grant-mapped memory and unmapped it; blkfront then
removed foreign access for that grant. The cost of unmapping scales
badly with the number of CPUs in Dom0. An experiment showed that when
Dom0 has 24 VCPUs, and guests are performing parallel I/O to a
ramdisk, the IPIs from performing unmap's is a bottleneck at 5 guests
(at which point 650,000 IOPS are being performed in total). If more
than 5 guests are used, the performance declines. By 10 guests, only
400,000 IOPS are being performed.
This patch improves performance by only unmapping when the connection
between blkfront and back is broken.
On startup blkfront notifies blkback that it is using persistent
grants, and blkback will do the same. If blkback is not capable of
persistent mapping, blkfront will still use the same grants, since it
is compatible with the previous protocol, and simplifies the code
complexity in blkfront.
To perform a read, in persistent mode, blkfront uses a separate pool
of pages that it maps to dom0. When a request comes in, blkfront
transmutes the request so that blkback will write into one of these
free pages. Blkback keeps note of which grefs it has already
mapped. When a new ring request comes to blkback, it looks to see if
it has already mapped that page. If so, it will not map it again. If
the page hasn't been previously mapped, it is mapped now, and a record
is kept of this mapping. Blkback proceeds as usual. When blkfront is
notified that blkback has completed a request, it memcpy's from the
shared memory, into the bvec supplied. A record that the {gref, page}
tuple is mapped, and not inflight is kept.
Writes are similar, except that the memcpy is peformed from the
supplied bvecs, into the shared pages, before the request is put onto
the ring.
Blkback stores a mapping of grefs=>{page mapped to by gref} in
a red-black tree. As the grefs are not known apriori, and provide no
guarantees on their ordering, we have to perform a search
through this tree to find the page, for every gref we receive. This
operation takes O(log n) time in the worst case. In blkfront grants
are stored using a single linked list.
The maximum number of grants that blkback will persistenly map is
currently set to RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, to
prevent a malicios guest from attempting a DoS, by supplying fresh
grefs, causing the Dom0 kernel to map excessively. If a guest
is using persistent grants and exceeds the maximum number of grants to
map persistenly the newly passed grefs will be mapped and unmaped.
Using this approach, we can have requests that mix persistent and
non-persistent grants, and we need to handle them correctly.
This allows us to set the maximum number of persistent grants to a
lower value than RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, although
setting it will lead to unpredictable performance.
In writing this patch, the question arrises as to if the additional
cost of performing memcpys in the guest (to/from the pool of granted
pages) outweigh the gains of not performing TLB shootdowns. The answer
to that question is `no'. There appears to be very little, if any
additional cost to the guest of using persistent grants. There is
perhaps a small saving, from the reduced number of hypercalls
performed in granting, and ending foreign access.
Signed-off-by: Oliver Chick <oliver.chick@citrix.com>
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[v1: Fixed up the misuse of bool as int]
2012-10-24 23:58:45 +07:00
|
|
|
|
|
|
|
struct persistent_gnt {
|
|
|
|
struct page *page;
|
|
|
|
grant_ref_t gnt;
|
|
|
|
grant_handle_t handle;
|
|
|
|
uint64_t dev_bus_addr;
|
|
|
|
struct rb_node node;
|
|
|
|
};
|
|
|
|
|
2011-05-13 03:47:48 +07:00
|
|
|
struct xen_blkif {
|
2011-04-15 05:25:47 +07:00
|
|
|
/* Unique identifier for this interface. */
|
2011-05-12 02:57:09 +07:00
|
|
|
domid_t domid;
|
|
|
|
unsigned int handle;
|
2011-04-15 05:25:47 +07:00
|
|
|
/* Physical parameters of the comms window. */
|
2011-05-12 02:57:09 +07:00
|
|
|
unsigned int irq;
|
2011-04-15 05:25:47 +07:00
|
|
|
/* Comms information. */
|
2011-05-12 02:57:09 +07:00
|
|
|
enum blkif_protocol blk_protocol;
|
|
|
|
union blkif_back_rings blk_rings;
|
2011-09-29 22:53:30 +07:00
|
|
|
void *blk_ring;
|
2011-04-15 05:25:47 +07:00
|
|
|
/* The VBD attached to this interface. */
|
2011-05-13 03:53:56 +07:00
|
|
|
struct xen_vbd vbd;
|
2011-04-15 05:25:47 +07:00
|
|
|
/* Back pointer to the backend_info. */
|
2011-05-12 02:57:09 +07:00
|
|
|
struct backend_info *be;
|
2011-04-15 05:25:47 +07:00
|
|
|
/* Private fields. */
|
2011-05-12 02:57:09 +07:00
|
|
|
spinlock_t blk_ring_lock;
|
|
|
|
atomic_t refcnt;
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-05-12 02:57:09 +07:00
|
|
|
wait_queue_head_t wq;
|
2011-10-10 11:42:22 +07:00
|
|
|
/* for barrier (drain) requests */
|
|
|
|
struct completion drain_complete;
|
|
|
|
atomic_t drain;
|
2011-04-15 04:05:23 +07:00
|
|
|
/* One thread per one blkif. */
|
2011-05-12 02:57:09 +07:00
|
|
|
struct task_struct *xenblkd;
|
|
|
|
unsigned int waiting_reqs;
|
2011-04-15 05:25:47 +07:00
|
|
|
|
xen/blkback: Persistent grant maps for xen blk drivers
This patch implements persistent grants for the xen-blk{front,back}
mechanism. The effect of this change is to reduce the number of unmap
operations performed, since they cause a (costly) TLB shootdown. This
allows the I/O performance to scale better when a large number of VMs
are performing I/O.
Previously, the blkfront driver was supplied a bvec[] from the request
queue. This was granted to dom0; dom0 performed the I/O and wrote
directly into the grant-mapped memory and unmapped it; blkfront then
removed foreign access for that grant. The cost of unmapping scales
badly with the number of CPUs in Dom0. An experiment showed that when
Dom0 has 24 VCPUs, and guests are performing parallel I/O to a
ramdisk, the IPIs from performing unmap's is a bottleneck at 5 guests
(at which point 650,000 IOPS are being performed in total). If more
than 5 guests are used, the performance declines. By 10 guests, only
400,000 IOPS are being performed.
This patch improves performance by only unmapping when the connection
between blkfront and back is broken.
On startup blkfront notifies blkback that it is using persistent
grants, and blkback will do the same. If blkback is not capable of
persistent mapping, blkfront will still use the same grants, since it
is compatible with the previous protocol, and simplifies the code
complexity in blkfront.
To perform a read, in persistent mode, blkfront uses a separate pool
of pages that it maps to dom0. When a request comes in, blkfront
transmutes the request so that blkback will write into one of these
free pages. Blkback keeps note of which grefs it has already
mapped. When a new ring request comes to blkback, it looks to see if
it has already mapped that page. If so, it will not map it again. If
the page hasn't been previously mapped, it is mapped now, and a record
is kept of this mapping. Blkback proceeds as usual. When blkfront is
notified that blkback has completed a request, it memcpy's from the
shared memory, into the bvec supplied. A record that the {gref, page}
tuple is mapped, and not inflight is kept.
Writes are similar, except that the memcpy is peformed from the
supplied bvecs, into the shared pages, before the request is put onto
the ring.
Blkback stores a mapping of grefs=>{page mapped to by gref} in
a red-black tree. As the grefs are not known apriori, and provide no
guarantees on their ordering, we have to perform a search
through this tree to find the page, for every gref we receive. This
operation takes O(log n) time in the worst case. In blkfront grants
are stored using a single linked list.
The maximum number of grants that blkback will persistenly map is
currently set to RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, to
prevent a malicios guest from attempting a DoS, by supplying fresh
grefs, causing the Dom0 kernel to map excessively. If a guest
is using persistent grants and exceeds the maximum number of grants to
map persistenly the newly passed grefs will be mapped and unmaped.
Using this approach, we can have requests that mix persistent and
non-persistent grants, and we need to handle them correctly.
This allows us to set the maximum number of persistent grants to a
lower value than RING_SIZE * BLKIF_MAX_SEGMENTS_PER_REQUEST, although
setting it will lead to unpredictable performance.
In writing this patch, the question arrises as to if the additional
cost of performing memcpys in the guest (to/from the pool of granted
pages) outweigh the gains of not performing TLB shootdowns. The answer
to that question is `no'. There appears to be very little, if any
additional cost to the guest of using persistent grants. There is
perhaps a small saving, from the reduced number of hypercalls
performed in granting, and ending foreign access.
Signed-off-by: Oliver Chick <oliver.chick@citrix.com>
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[v1: Fixed up the misuse of bool as int]
2012-10-24 23:58:45 +07:00
|
|
|
/* tree to store persistent grants */
|
|
|
|
struct rb_root persistent_gnts;
|
|
|
|
unsigned int persistent_gnt_c;
|
|
|
|
|
2011-04-15 05:25:47 +07:00
|
|
|
/* statistics */
|
2011-05-12 02:57:09 +07:00
|
|
|
unsigned long st_print;
|
|
|
|
int st_rd_req;
|
|
|
|
int st_wr_req;
|
|
|
|
int st_oo_req;
|
|
|
|
int st_f_req;
|
2011-09-01 17:39:10 +07:00
|
|
|
int st_ds_req;
|
2011-05-12 02:57:09 +07:00
|
|
|
int st_rd_sect;
|
|
|
|
int st_wr_sect;
|
|
|
|
|
|
|
|
wait_queue_head_t waiting_to_free;
|
2011-04-15 04:21:50 +07:00
|
|
|
};
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-04-20 22:21:43 +07:00
|
|
|
|
|
|
|
#define vbd_sz(_v) ((_v)->bdev->bd_part ? \
|
|
|
|
(_v)->bdev->bd_part->nr_sects : \
|
|
|
|
get_capacity((_v)->bdev->bd_disk))
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-04-20 22:50:43 +07:00
|
|
|
#define xen_blkif_get(_b) (atomic_inc(&(_b)->refcnt))
|
|
|
|
#define xen_blkif_put(_b) \
|
2011-04-15 05:25:47 +07:00
|
|
|
do { \
|
|
|
|
if (atomic_dec_and_test(&(_b)->refcnt)) \
|
|
|
|
wake_up(&(_b)->waiting_to_free);\
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
struct phys_req {
|
2011-05-12 02:57:09 +07:00
|
|
|
unsigned short dev;
|
2011-09-01 17:39:10 +07:00
|
|
|
blkif_sector_t nr_sects;
|
2011-05-12 02:57:09 +07:00
|
|
|
struct block_device *bdev;
|
|
|
|
blkif_sector_t sector_number;
|
2011-04-15 05:25:47 +07:00
|
|
|
};
|
2011-04-20 22:50:43 +07:00
|
|
|
int xen_blkif_interface_init(void);
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-04-20 22:50:43 +07:00
|
|
|
int xen_blkif_xenbus_init(void);
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-04-20 22:50:43 +07:00
|
|
|
irqreturn_t xen_blkif_be_int(int irq, void *dev_id);
|
|
|
|
int xen_blkif_schedule(void *arg);
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-05-05 04:07:27 +07:00
|
|
|
int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
|
|
|
|
struct backend_info *be, int state);
|
2011-04-15 05:25:47 +07:00
|
|
|
|
2011-10-10 11:42:22 +07:00
|
|
|
int xen_blkbk_barrier(struct xenbus_transaction xbt,
|
|
|
|
struct backend_info *be, int state);
|
2011-04-20 22:50:43 +07:00
|
|
|
struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be);
|
2010-03-19 05:35:05 +07:00
|
|
|
|
2011-05-13 03:23:06 +07:00
|
|
|
static inline void blkif_get_x86_32_req(struct blkif_request *dst,
|
2011-05-12 03:23:39 +07:00
|
|
|
struct blkif_x86_32_request *src)
|
|
|
|
{
|
|
|
|
int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
|
|
|
|
dst->operation = src->operation;
|
2011-09-01 17:39:10 +07:00
|
|
|
switch (src->operation) {
|
|
|
|
case BLKIF_OP_READ:
|
|
|
|
case BLKIF_OP_WRITE:
|
|
|
|
case BLKIF_OP_WRITE_BARRIER:
|
|
|
|
case BLKIF_OP_FLUSH_DISKCACHE:
|
2011-10-12 23:12:36 +07:00
|
|
|
dst->u.rw.nr_segments = src->u.rw.nr_segments;
|
|
|
|
dst->u.rw.handle = src->u.rw.handle;
|
|
|
|
dst->u.rw.id = src->u.rw.id;
|
2011-09-01 17:39:10 +07:00
|
|
|
dst->u.rw.sector_number = src->u.rw.sector_number;
|
|
|
|
barrier();
|
2011-10-12 23:12:36 +07:00
|
|
|
if (n > dst->u.rw.nr_segments)
|
|
|
|
n = dst->u.rw.nr_segments;
|
2011-09-01 17:39:10 +07:00
|
|
|
for (i = 0; i < n; i++)
|
|
|
|
dst->u.rw.seg[i] = src->u.rw.seg[i];
|
|
|
|
break;
|
|
|
|
case BLKIF_OP_DISCARD:
|
2011-10-13 03:23:30 +07:00
|
|
|
dst->u.discard.flag = src->u.discard.flag;
|
2012-05-26 03:11:09 +07:00
|
|
|
dst->u.discard.id = src->u.discard.id;
|
2011-09-01 17:39:10 +07:00
|
|
|
dst->u.discard.sector_number = src->u.discard.sector_number;
|
|
|
|
dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
2011-05-12 03:23:39 +07:00
|
|
|
}
|
|
|
|
|
2011-05-13 03:23:06 +07:00
|
|
|
static inline void blkif_get_x86_64_req(struct blkif_request *dst,
|
2011-05-12 03:23:39 +07:00
|
|
|
struct blkif_x86_64_request *src)
|
|
|
|
{
|
|
|
|
int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
|
|
|
|
dst->operation = src->operation;
|
2011-09-01 17:39:10 +07:00
|
|
|
switch (src->operation) {
|
|
|
|
case BLKIF_OP_READ:
|
|
|
|
case BLKIF_OP_WRITE:
|
|
|
|
case BLKIF_OP_WRITE_BARRIER:
|
|
|
|
case BLKIF_OP_FLUSH_DISKCACHE:
|
2011-10-12 23:12:36 +07:00
|
|
|
dst->u.rw.nr_segments = src->u.rw.nr_segments;
|
|
|
|
dst->u.rw.handle = src->u.rw.handle;
|
|
|
|
dst->u.rw.id = src->u.rw.id;
|
2011-09-01 17:39:10 +07:00
|
|
|
dst->u.rw.sector_number = src->u.rw.sector_number;
|
|
|
|
barrier();
|
2011-10-12 23:12:36 +07:00
|
|
|
if (n > dst->u.rw.nr_segments)
|
|
|
|
n = dst->u.rw.nr_segments;
|
2011-09-01 17:39:10 +07:00
|
|
|
for (i = 0; i < n; i++)
|
|
|
|
dst->u.rw.seg[i] = src->u.rw.seg[i];
|
|
|
|
break;
|
|
|
|
case BLKIF_OP_DISCARD:
|
2011-10-13 03:23:30 +07:00
|
|
|
dst->u.discard.flag = src->u.discard.flag;
|
2012-05-26 03:11:09 +07:00
|
|
|
dst->u.discard.id = src->u.discard.id;
|
2011-09-01 17:39:10 +07:00
|
|
|
dst->u.discard.sector_number = src->u.discard.sector_number;
|
|
|
|
dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
2011-05-12 03:23:39 +07:00
|
|
|
}
|
|
|
|
|
2011-05-13 03:58:21 +07:00
|
|
|
#endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */
|