aboutsummaryrefslogtreecommitdiffstats
path: root/block
diff options
context:
space:
mode:
authorTejun Heo <tj@kernel.org>2009-07-03 18:13:18 -0400
committerTejun Heo <tj@kernel.org>2009-07-03 18:13:18 -0400
commitc43768cbb7655ea5ff782ae250f6e2ef4297cf98 (patch)
tree3982e41dde3eecaa3739a5d1a8ed18d04bd74f01 /block
parent1a8dd307cc0a2119be4e578c517795464e6dabba (diff)
parent746a99a5af60ee676afa2ba469ccd1373493c7e7 (diff)
Merge branch 'master' into for-next
Pull linus#master to merge PER_CPU_DEF_ATTRIBUTES and alpha build fix changes. As alpha in percpu tree uses 'weak' attribute instead of inline assembly, there's no need for __used attribute. Conflicts: arch/alpha/include/asm/percpu.h arch/mn10300/kernel/vmlinux.lds.S include/linux/percpu-defs.h
Diffstat (limited to 'block')
-rw-r--r--block/Makefile2
-rw-r--r--block/blk-core.c14
-rw-r--r--block/bsg.c5
-rw-r--r--block/cfq-iosched.c176
-rw-r--r--block/cmd-filter.c233
-rw-r--r--block/scsi_ioctl.c43
6 files changed, 139 insertions, 334 deletions
diff --git a/block/Makefile b/block/Makefile
index e9fa4dd690f2..6c54ed0ff755 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -5,7 +5,7 @@
5obj-$(CONFIG_BLOCK) := elevator.o blk-core.o blk-tag.o blk-sysfs.o \ 5obj-$(CONFIG_BLOCK) := elevator.o blk-core.o blk-tag.o blk-sysfs.o \
6 blk-barrier.o blk-settings.o blk-ioc.o blk-map.o \ 6 blk-barrier.o blk-settings.o blk-ioc.o blk-map.o \
7 blk-exec.o blk-merge.o blk-softirq.o blk-timeout.o \ 7 blk-exec.o blk-merge.o blk-softirq.o blk-timeout.o \
8 ioctl.o genhd.o scsi_ioctl.o cmd-filter.o 8 ioctl.o genhd.o scsi_ioctl.o
9 9
10obj-$(CONFIG_BLK_DEV_BSG) += bsg.o 10obj-$(CONFIG_BLK_DEV_BSG) += bsg.o
11obj-$(CONFIG_IOSCHED_NOOP) += noop-iosched.o 11obj-$(CONFIG_IOSCHED_NOOP) += noop-iosched.o
diff --git a/block/blk-core.c b/block/blk-core.c
index b06cf5c2a829..4b45435c6eaf 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -595,8 +595,6 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
595 595
596 q->sg_reserved_size = INT_MAX; 596 q->sg_reserved_size = INT_MAX;
597 597
598 blk_set_cmd_filter_defaults(&q->cmd_filter);
599
600 /* 598 /*
601 * all done 599 * all done
602 */ 600 */
@@ -1172,6 +1170,11 @@ static int __make_request(struct request_queue *q, struct bio *bio)
1172 const int unplug = bio_unplug(bio); 1170 const int unplug = bio_unplug(bio);
1173 int rw_flags; 1171 int rw_flags;
1174 1172
1173 if (bio_barrier(bio) && bio_has_data(bio) &&
1174 (q->next_ordered == QUEUE_ORDERED_NONE)) {
1175 bio_endio(bio, -EOPNOTSUPP);
1176 return 0;
1177 }
1175 /* 1178 /*
1176 * low level driver can indicate that it wants pages above a 1179 * low level driver can indicate that it wants pages above a
1177 * certain limit bounced to low memory (ie for highmem, or even 1180 * certain limit bounced to low memory (ie for highmem, or even
@@ -1472,11 +1475,6 @@ static inline void __generic_make_request(struct bio *bio)
1472 err = -EOPNOTSUPP; 1475 err = -EOPNOTSUPP;
1473 goto end_io; 1476 goto end_io;
1474 } 1477 }
1475 if (bio_barrier(bio) && bio_has_data(bio) &&
1476 (q->next_ordered == QUEUE_ORDERED_NONE)) {
1477 err = -EOPNOTSUPP;
1478 goto end_io;
1479 }
1480 1478
1481 ret = q->make_request_fn(q, bio); 1479 ret = q->make_request_fn(q, bio);
1482 } while (ret); 1480 } while (ret);
@@ -2365,7 +2363,7 @@ int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
2365 __bio_clone(bio, bio_src); 2363 __bio_clone(bio, bio_src);
2366 2364
2367 if (bio_integrity(bio_src) && 2365 if (bio_integrity(bio_src) &&
2368 bio_integrity_clone(bio, bio_src, gfp_mask)) 2366 bio_integrity_clone(bio, bio_src, gfp_mask, bs))
2369 goto free_and_out; 2367 goto free_and_out;
2370 2368
2371 if (bio_ctr && bio_ctr(bio, bio_src, data)) 2369 if (bio_ctr && bio_ctr(bio, bio_src, data))
diff --git a/block/bsg.c b/block/bsg.c
index 54106f052f70..5f184bb3ff9e 100644
--- a/block/bsg.c
+++ b/block/bsg.c
@@ -186,7 +186,7 @@ static int blk_fill_sgv4_hdr_rq(struct request_queue *q, struct request *rq,
186 return -EFAULT; 186 return -EFAULT;
187 187
188 if (hdr->subprotocol == BSG_SUB_PROTOCOL_SCSI_CMD) { 188 if (hdr->subprotocol == BSG_SUB_PROTOCOL_SCSI_CMD) {
189 if (blk_verify_command(&q->cmd_filter, rq->cmd, has_write_perm)) 189 if (blk_verify_command(rq->cmd, has_write_perm))
190 return -EPERM; 190 return -EPERM;
191 } else if (!capable(CAP_SYS_RAWIO)) 191 } else if (!capable(CAP_SYS_RAWIO))
192 return -EPERM; 192 return -EPERM;
@@ -315,7 +315,6 @@ out:
315 blk_put_request(rq); 315 blk_put_request(rq);
316 if (next_rq) { 316 if (next_rq) {
317 blk_rq_unmap_user(next_rq->bio); 317 blk_rq_unmap_user(next_rq->bio);
318 next_rq->bio = NULL;
319 blk_put_request(next_rq); 318 blk_put_request(next_rq);
320 } 319 }
321 return ERR_PTR(ret); 320 return ERR_PTR(ret);
@@ -449,7 +448,6 @@ static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
449 hdr->dout_resid = rq->resid_len; 448 hdr->dout_resid = rq->resid_len;
450 hdr->din_resid = rq->next_rq->resid_len; 449 hdr->din_resid = rq->next_rq->resid_len;
451 blk_rq_unmap_user(bidi_bio); 450 blk_rq_unmap_user(bidi_bio);
452 rq->next_rq->bio = NULL;
453 blk_put_request(rq->next_rq); 451 blk_put_request(rq->next_rq);
454 } else if (rq_data_dir(rq) == READ) 452 } else if (rq_data_dir(rq) == READ)
455 hdr->din_resid = rq->resid_len; 453 hdr->din_resid = rq->resid_len;
@@ -468,7 +466,6 @@ static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
468 blk_rq_unmap_user(bio); 466 blk_rq_unmap_user(bio);
469 if (rq->cmd != rq->__cmd) 467 if (rq->cmd != rq->__cmd)
470 kfree(rq->cmd); 468 kfree(rq->cmd);
471 rq->bio = NULL;
472 blk_put_request(rq); 469 blk_put_request(rq);
473 470
474 return ret; 471 return ret;
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 0f1cc7d3855e..85208dd1d05b 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -71,6 +71,51 @@ struct cfq_rb_root {
71#define CFQ_RB_ROOT (struct cfq_rb_root) { RB_ROOT, NULL, } 71#define CFQ_RB_ROOT (struct cfq_rb_root) { RB_ROOT, NULL, }
72 72
73/* 73/*
74 * Per process-grouping structure
75 */
76struct cfq_queue {
77 /* reference count */
78 atomic_t ref;
79 /* various state flags, see below */
80 unsigned int flags;
81 /* parent cfq_data */
82 struct cfq_data *cfqd;
83 /* service_tree member */
84 struct rb_node rb_node;
85 /* service_tree key */
86 unsigned long rb_key;
87 /* prio tree member */
88 struct rb_node p_node;
89 /* prio tree root we belong to, if any */
90 struct rb_root *p_root;
91 /* sorted list of pending requests */
92 struct rb_root sort_list;
93 /* if fifo isn't expired, next request to serve */
94 struct request *next_rq;
95 /* requests queued in sort_list */
96 int queued[2];
97 /* currently allocated requests */
98 int allocated[2];
99 /* fifo list of requests in sort_list */
100 struct list_head fifo;
101
102 unsigned long slice_end;
103 long slice_resid;
104 unsigned int slice_dispatch;
105
106 /* pending metadata requests */
107 int meta_pending;
108 /* number of requests that are on the dispatch list or inside driver */
109 int dispatched;
110
111 /* io prio of this group */
112 unsigned short ioprio, org_ioprio;
113 unsigned short ioprio_class, org_ioprio_class;
114
115 pid_t pid;
116};
117
118/*
74 * Per block device queue structure 119 * Per block device queue structure
75 */ 120 */
76struct cfq_data { 121struct cfq_data {
@@ -135,51 +180,11 @@ struct cfq_data {
135 unsigned int cfq_slice_idle; 180 unsigned int cfq_slice_idle;
136 181
137 struct list_head cic_list; 182 struct list_head cic_list;
138};
139
140/*
141 * Per process-grouping structure
142 */
143struct cfq_queue {
144 /* reference count */
145 atomic_t ref;
146 /* various state flags, see below */
147 unsigned int flags;
148 /* parent cfq_data */
149 struct cfq_data *cfqd;
150 /* service_tree member */
151 struct rb_node rb_node;
152 /* service_tree key */
153 unsigned long rb_key;
154 /* prio tree member */
155 struct rb_node p_node;
156 /* prio tree root we belong to, if any */
157 struct rb_root *p_root;
158 /* sorted list of pending requests */
159 struct rb_root sort_list;
160 /* if fifo isn't expired, next request to serve */
161 struct request *next_rq;
162 /* requests queued in sort_list */
163 int queued[2];
164 /* currently allocated requests */
165 int allocated[2];
166 /* fifo list of requests in sort_list */
167 struct list_head fifo;
168 183
169 unsigned long slice_end; 184 /*
170 long slice_resid; 185 * Fallback dummy cfqq for extreme OOM conditions
171 unsigned int slice_dispatch; 186 */
172 187 struct cfq_queue oom_cfqq;
173 /* pending metadata requests */
174 int meta_pending;
175 /* number of requests that are on the dispatch list or inside driver */
176 int dispatched;
177
178 /* io prio of this group */
179 unsigned short ioprio, org_ioprio;
180 unsigned short ioprio_class, org_ioprio_class;
181
182 pid_t pid;
183}; 188};
184 189
185enum cfqq_state_flags { 190enum cfqq_state_flags {
@@ -1641,6 +1646,26 @@ static void cfq_ioc_set_ioprio(struct io_context *ioc)
1641 ioc->ioprio_changed = 0; 1646 ioc->ioprio_changed = 0;
1642} 1647}
1643 1648
1649static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1650 pid_t pid, int is_sync)
1651{
1652 RB_CLEAR_NODE(&cfqq->rb_node);
1653 RB_CLEAR_NODE(&cfqq->p_node);
1654 INIT_LIST_HEAD(&cfqq->fifo);
1655
1656 atomic_set(&cfqq->ref, 0);
1657 cfqq->cfqd = cfqd;
1658
1659 cfq_mark_cfqq_prio_changed(cfqq);
1660
1661 if (is_sync) {
1662 if (!cfq_class_idle(cfqq))
1663 cfq_mark_cfqq_idle_window(cfqq);
1664 cfq_mark_cfqq_sync(cfqq);
1665 }
1666 cfqq->pid = pid;
1667}
1668
1644static struct cfq_queue * 1669static struct cfq_queue *
1645cfq_find_alloc_queue(struct cfq_data *cfqd, int is_sync, 1670cfq_find_alloc_queue(struct cfq_data *cfqd, int is_sync,
1646 struct io_context *ioc, gfp_t gfp_mask) 1671 struct io_context *ioc, gfp_t gfp_mask)
@@ -1653,56 +1678,40 @@ retry:
1653 /* cic always exists here */ 1678 /* cic always exists here */
1654 cfqq = cic_to_cfqq(cic, is_sync); 1679 cfqq = cic_to_cfqq(cic, is_sync);
1655 1680
1656 if (!cfqq) { 1681 /*
1682 * Always try a new alloc if we fell back to the OOM cfqq
1683 * originally, since it should just be a temporary situation.
1684 */
1685 if (!cfqq || cfqq == &cfqd->oom_cfqq) {
1686 cfqq = NULL;
1657 if (new_cfqq) { 1687 if (new_cfqq) {
1658 cfqq = new_cfqq; 1688 cfqq = new_cfqq;
1659 new_cfqq = NULL; 1689 new_cfqq = NULL;
1660 } else if (gfp_mask & __GFP_WAIT) { 1690 } else if (gfp_mask & __GFP_WAIT) {
1661 /*
1662 * Inform the allocator of the fact that we will
1663 * just repeat this allocation if it fails, to allow
1664 * the allocator to do whatever it needs to attempt to
1665 * free memory.
1666 */
1667 spin_unlock_irq(cfqd->queue->queue_lock); 1691 spin_unlock_irq(cfqd->queue->queue_lock);
1668 new_cfqq = kmem_cache_alloc_node(cfq_pool, 1692 new_cfqq = kmem_cache_alloc_node(cfq_pool,
1669 gfp_mask | __GFP_NOFAIL | __GFP_ZERO, 1693 gfp_mask | __GFP_ZERO,
1670 cfqd->queue->node); 1694 cfqd->queue->node);
1671 spin_lock_irq(cfqd->queue->queue_lock); 1695 spin_lock_irq(cfqd->queue->queue_lock);
1672 goto retry; 1696 if (new_cfqq)
1697 goto retry;
1673 } else { 1698 } else {
1674 cfqq = kmem_cache_alloc_node(cfq_pool, 1699 cfqq = kmem_cache_alloc_node(cfq_pool,
1675 gfp_mask | __GFP_ZERO, 1700 gfp_mask | __GFP_ZERO,
1676 cfqd->queue->node); 1701 cfqd->queue->node);
1677 if (!cfqq)
1678 goto out;
1679 } 1702 }
1680 1703
1681 RB_CLEAR_NODE(&cfqq->rb_node); 1704 if (cfqq) {
1682 RB_CLEAR_NODE(&cfqq->p_node); 1705 cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
1683 INIT_LIST_HEAD(&cfqq->fifo); 1706 cfq_init_prio_data(cfqq, ioc);
1684 1707 cfq_log_cfqq(cfqd, cfqq, "alloced");
1685 atomic_set(&cfqq->ref, 0); 1708 } else
1686 cfqq->cfqd = cfqd; 1709 cfqq = &cfqd->oom_cfqq;
1687
1688 cfq_mark_cfqq_prio_changed(cfqq);
1689
1690 cfq_init_prio_data(cfqq, ioc);
1691
1692 if (is_sync) {
1693 if (!cfq_class_idle(cfqq))
1694 cfq_mark_cfqq_idle_window(cfqq);
1695 cfq_mark_cfqq_sync(cfqq);
1696 }
1697 cfqq->pid = current->pid;
1698 cfq_log_cfqq(cfqd, cfqq, "alloced");
1699 } 1710 }
1700 1711
1701 if (new_cfqq) 1712 if (new_cfqq)
1702 kmem_cache_free(cfq_pool, new_cfqq); 1713 kmem_cache_free(cfq_pool, new_cfqq);
1703 1714
1704out:
1705 WARN_ON((gfp_mask & __GFP_WAIT) && !cfqq);
1706 return cfqq; 1715 return cfqq;
1707} 1716}
1708 1717
@@ -1735,11 +1744,8 @@ cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct io_context *ioc,
1735 cfqq = *async_cfqq; 1744 cfqq = *async_cfqq;
1736 } 1745 }
1737 1746
1738 if (!cfqq) { 1747 if (!cfqq)
1739 cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask); 1748 cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask);
1740 if (!cfqq)
1741 return NULL;
1742 }
1743 1749
1744 /* 1750 /*
1745 * pin the queue now that it's allocated, scheduler exit will prune it 1751 * pin the queue now that it's allocated, scheduler exit will prune it
@@ -2307,10 +2313,6 @@ cfq_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask)
2307 cfqq = cic_to_cfqq(cic, is_sync); 2313 cfqq = cic_to_cfqq(cic, is_sync);
2308 if (!cfqq) { 2314 if (!cfqq) {
2309 cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask); 2315 cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask);
2310
2311 if (!cfqq)
2312 goto queue_fail;
2313
2314 cic_set_cfqq(cic, cfqq, is_sync); 2316 cic_set_cfqq(cic, cfqq, is_sync);
2315 } 2317 }
2316 2318
@@ -2465,6 +2467,14 @@ static void *cfq_init_queue(struct request_queue *q)
2465 for (i = 0; i < CFQ_PRIO_LISTS; i++) 2467 for (i = 0; i < CFQ_PRIO_LISTS; i++)
2466 cfqd->prio_trees[i] = RB_ROOT; 2468 cfqd->prio_trees[i] = RB_ROOT;
2467 2469
2470 /*
2471 * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues.
2472 * Grab a permanent reference to it, so that the normal code flow
2473 * will not attempt to free it.
2474 */
2475 cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
2476 atomic_inc(&cfqd->oom_cfqq.ref);
2477
2468 INIT_LIST_HEAD(&cfqd->cic_list); 2478 INIT_LIST_HEAD(&cfqd->cic_list);
2469 2479
2470 cfqd->queue = q; 2480 cfqd->queue = q;
diff --git a/block/cmd-filter.c b/block/cmd-filter.c
deleted file mode 100644
index 572bbc2f900d..000000000000
--- a/block/cmd-filter.c
+++ /dev/null
@@ -1,233 +0,0 @@
1/*
2 * Copyright 2004 Peter M. Jones <pjones@redhat.com>
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 *
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public Licens
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
17 *
18 */
19
20#include <linux/list.h>
21#include <linux/genhd.h>
22#include <linux/spinlock.h>
23#include <linux/capability.h>
24#include <linux/bitops.h>
25#include <linux/blkdev.h>
26
27#include <scsi/scsi.h>
28#include <linux/cdrom.h>
29
30int blk_verify_command(struct blk_cmd_filter *filter,
31 unsigned char *cmd, fmode_t has_write_perm)
32{
33 /* root can do any command. */
34 if (capable(CAP_SYS_RAWIO))
35 return 0;
36
37 /* if there's no filter set, assume we're filtering everything out */
38 if (!filter)
39 return -EPERM;
40
41 /* Anybody who can open the device can do a read-safe command */
42 if (test_bit(cmd[0], filter->read_ok))
43 return 0;
44
45 /* Write-safe commands require a writable open */
46 if (test_bit(cmd[0], filter->write_ok) && has_write_perm)
47 return 0;
48
49 return -EPERM;
50}
51EXPORT_SYMBOL(blk_verify_command);
52
53#if 0
54/* and now, the sysfs stuff */
55static ssize_t rcf_cmds_show(struct blk_cmd_filter *filter, char *page,
56 int rw)
57{
58 char *npage = page;
59 unsigned long *okbits;
60 int i;
61
62 if (rw == READ)
63 okbits = filter->read_ok;
64 else
65 okbits = filter->write_ok;
66
67 for (i = 0; i < BLK_SCSI_MAX_CMDS; i++) {
68 if (test_bit(i, okbits)) {
69 npage += sprintf(npage, "0x%02x", i);
70 if (i < BLK_SCSI_MAX_CMDS - 1)
71 sprintf(npage++, " ");
72 }
73 }
74
75 if (npage != page)
76 npage += sprintf(npage, "\n");
77
78 return npage - page;
79}
80
81static ssize_t rcf_readcmds_show(struct blk_cmd_filter *filter, char *page)
82{
83 return rcf_cmds_show(filter, page, READ);
84}
85
86static ssize_t rcf_writecmds_show(struct blk_cmd_filter *filter,
87 char *page)
88{
89 return rcf_cmds_show(filter, page, WRITE);
90}
91
92static ssize_t rcf_cmds_store(struct blk_cmd_filter *filter,
93 const char *page, size_t count, int rw)
94{
95 unsigned long okbits[BLK_SCSI_CMD_PER_LONG], *target_okbits;
96 int cmd, set;
97 char *p, *status;
98
99 if (rw == READ) {
100 memcpy(&okbits, filter->read_ok, sizeof(okbits));
101 target_okbits = filter->read_ok;
102 } else {
103 memcpy(&okbits, filter->write_ok, sizeof(okbits));
104 target_okbits = filter->write_ok;
105 }
106
107 while ((p = strsep((char **)&page, " ")) != NULL) {
108 set = 1;
109
110 if (p[0] == '+') {
111 p++;
112 } else if (p[0] == '-') {
113 set = 0;
114 p++;
115 }
116
117 cmd = simple_strtol(p, &status, 16);
118
119 /* either of these cases means invalid input, so do nothing. */
120 if ((status == p) || cmd >= BLK_SCSI_MAX_CMDS)
121 return -EINVAL;
122
123 if (set)
124 __set_bit(cmd, okbits);
125 else
126 __clear_bit(cmd, okbits);
127 }
128
129 memcpy(target_okbits, okbits, sizeof(okbits));
130 return count;
131}
132
133static ssize_t rcf_readcmds_store(struct blk_cmd_filter *filter,
134 const char *page, size_t count)
135{
136 return rcf_cmds_store(filter, page, count, READ);
137}
138
139static ssize_t rcf_writecmds_store(struct blk_cmd_filter *filter,
140 const char *page, size_t count)
141{
142 return rcf_cmds_store(filter, page, count, WRITE);
143}
144
145struct rcf_sysfs_entry {
146 struct attribute attr;
147 ssize_t (*show)(struct blk_cmd_filter *, char *);
148 ssize_t (*store)(struct blk_cmd_filter *, const char *, size_t);
149};
150
151static struct rcf_sysfs_entry rcf_readcmds_entry = {
152 .attr = { .name = "read_table", .mode = S_IRUGO | S_IWUSR },
153 .show = rcf_readcmds_show,
154 .store = rcf_readcmds_store,
155};
156
157static struct rcf_sysfs_entry rcf_writecmds_entry = {
158 .attr = {.name = "write_table", .mode = S_IRUGO | S_IWUSR },
159 .show = rcf_writecmds_show,
160 .store = rcf_writecmds_store,
161};
162
163static struct attribute *default_attrs[] = {
164 &rcf_readcmds_entry.attr,
165 &rcf_writecmds_entry.attr,
166 NULL,
167};
168
169#define to_rcf(atr) container_of((atr), struct rcf_sysfs_entry, attr)
170
171static ssize_t
172rcf_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
173{
174 struct rcf_sysfs_entry *entry = to_rcf(attr);
175 struct blk_cmd_filter *filter;
176
177 filter = container_of(kobj, struct blk_cmd_filter, kobj);
178 if (entry->show)
179 return entry->show(filter, page);
180
181 return 0;
182}
183
184static ssize_t
185rcf_attr_store(struct kobject *kobj, struct attribute *attr,
186 const char *page, size_t length)
187{
188 struct rcf_sysfs_entry *entry = to_rcf(attr);
189 struct blk_cmd_filter *filter;
190
191 if (!capable(CAP_SYS_RAWIO))
192 return -EPERM;
193
194 if (!entry->store)
195 return -EINVAL;
196
197 filter = container_of(kobj, struct blk_cmd_filter, kobj);
198 return entry->store(filter, page, length);
199}
200
201static struct sysfs_ops rcf_sysfs_ops = {
202 .show = rcf_attr_show,
203 .store = rcf_attr_store,
204};
205
206static struct kobj_type rcf_ktype = {
207 .sysfs_ops = &rcf_sysfs_ops,
208 .default_attrs = default_attrs,
209};
210
211int blk_register_filter(struct gendisk *disk)
212{
213 int ret;
214 struct blk_cmd_filter *filter = &disk->queue->cmd_filter;
215
216 ret = kobject_init_and_add(&filter->kobj, &rcf_ktype,
217 &disk_to_dev(disk)->kobj,
218 "%s", "cmd_filter");
219 if (ret < 0)
220 return ret;
221
222 return 0;
223}
224EXPORT_SYMBOL(blk_register_filter);
225
226void blk_unregister_filter(struct gendisk *disk)
227{
228 struct blk_cmd_filter *filter = &disk->queue->cmd_filter;
229
230 kobject_put(&filter->kobj);
231}
232EXPORT_SYMBOL(blk_unregister_filter);
233#endif
diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index 5f8e798ede4e..f0e0ce0a607d 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -32,6 +32,11 @@
32#include <scsi/scsi_ioctl.h> 32#include <scsi/scsi_ioctl.h>
33#include <scsi/scsi_cmnd.h> 33#include <scsi/scsi_cmnd.h>
34 34
35struct blk_cmd_filter {
36 unsigned long read_ok[BLK_SCSI_CMD_PER_LONG];
37 unsigned long write_ok[BLK_SCSI_CMD_PER_LONG];
38} blk_default_cmd_filter;
39
35/* Command group 3 is reserved and should never be used. */ 40/* Command group 3 is reserved and should never be used. */
36const unsigned char scsi_command_size_tbl[8] = 41const unsigned char scsi_command_size_tbl[8] =
37{ 42{
@@ -105,7 +110,7 @@ static int sg_emulated_host(struct request_queue *q, int __user *p)
105 return put_user(1, p); 110 return put_user(1, p);
106} 111}
107 112
108void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter) 113static void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter)
109{ 114{
110 /* Basic read-only commands */ 115 /* Basic read-only commands */
111 __set_bit(TEST_UNIT_READY, filter->read_ok); 116 __set_bit(TEST_UNIT_READY, filter->read_ok);
@@ -187,14 +192,37 @@ void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter)
187 __set_bit(GPCMD_SET_STREAMING, filter->write_ok); 192 __set_bit(GPCMD_SET_STREAMING, filter->write_ok);
188 __set_bit(GPCMD_SET_READ_AHEAD, filter->write_ok); 193 __set_bit(GPCMD_SET_READ_AHEAD, filter->write_ok);
189} 194}
190EXPORT_SYMBOL_GPL(blk_set_cmd_filter_defaults); 195
196int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm)
197{
198 struct blk_cmd_filter *filter = &blk_default_cmd_filter;
199
200 /* root can do any command. */
201 if (capable(CAP_SYS_RAWIO))
202 return 0;
203
204 /* if there's no filter set, assume we're filtering everything out */
205 if (!filter)
206 return -EPERM;
207
208 /* Anybody who can open the device can do a read-safe command */
209 if (test_bit(cmd[0], filter->read_ok))
210 return 0;
211
212 /* Write-safe commands require a writable open */
213 if (test_bit(cmd[0], filter->write_ok) && has_write_perm)
214 return 0;
215
216 return -EPERM;
217}
218EXPORT_SYMBOL(blk_verify_command);
191 219
192static int blk_fill_sghdr_rq(struct request_queue *q, struct request *rq, 220static int blk_fill_sghdr_rq(struct request_queue *q, struct request *rq,
193 struct sg_io_hdr *hdr, fmode_t mode) 221 struct sg_io_hdr *hdr, fmode_t mode)
194{ 222{
195 if (copy_from_user(rq->cmd, hdr->cmdp, hdr->cmd_len)) 223 if (copy_from_user(rq->cmd, hdr->cmdp, hdr->cmd_len))
196 return -EFAULT; 224 return -EFAULT;
197 if (blk_verify_command(&q->cmd_filter, rq->cmd, mode & FMODE_WRITE)) 225 if (blk_verify_command(rq->cmd, mode & FMODE_WRITE))
198 return -EPERM; 226 return -EPERM;
199 227
200 /* 228 /*
@@ -427,7 +455,7 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
427 if (in_len && copy_from_user(buffer, sic->data + cmdlen, in_len)) 455 if (in_len && copy_from_user(buffer, sic->data + cmdlen, in_len))
428 goto error; 456 goto error;
429 457
430 err = blk_verify_command(&q->cmd_filter, rq->cmd, mode & FMODE_WRITE); 458 err = blk_verify_command(rq->cmd, mode & FMODE_WRITE);
431 if (err) 459 if (err)
432 goto error; 460 goto error;
433 461
@@ -645,5 +673,10 @@ int scsi_cmd_ioctl(struct request_queue *q, struct gendisk *bd_disk, fmode_t mod
645 blk_put_queue(q); 673 blk_put_queue(q);
646 return err; 674 return err;
647} 675}
648
649EXPORT_SYMBOL(scsi_cmd_ioctl); 676EXPORT_SYMBOL(scsi_cmd_ioctl);
677
678int __init blk_scsi_ioctl_init(void)
679{
680 blk_set_cmd_filter_defaults(&blk_default_cmd_filter);
681 return 0;
682}