diff options
Diffstat (limited to 'drivers/xen')
-rw-r--r-- | drivers/xen/events.c | 101 | ||||
-rw-r--r-- | drivers/xen/xenbus/xenbus_probe.c | 4 | ||||
-rw-r--r-- | drivers/xen/xenfs/Makefile | 3 | ||||
-rw-r--r-- | drivers/xen/xenfs/privcmd.c | 404 | ||||
-rw-r--r-- | drivers/xen/xenfs/super.c | 95 | ||||
-rw-r--r-- | drivers/xen/xenfs/xenfs.h | 3 | ||||
-rw-r--r-- | drivers/xen/xenfs/xenstored.c | 68 |
7 files changed, 639 insertions, 39 deletions
diff --git a/drivers/xen/events.c b/drivers/xen/events.c index 7d24b0d94ed4..347f17edad77 100644 --- a/drivers/xen/events.c +++ b/drivers/xen/events.c | |||
@@ -261,7 +261,7 @@ static void init_evtchn_cpu_bindings(void) | |||
261 | } | 261 | } |
262 | #endif | 262 | #endif |
263 | 263 | ||
264 | memset(cpu_evtchn_mask(0), ~0, sizeof(cpu_evtchn_mask(0))); | 264 | memset(cpu_evtchn_mask(0), ~0, sizeof(struct cpu_evtchn_s)); |
265 | } | 265 | } |
266 | 266 | ||
267 | static inline void clear_evtchn(int port) | 267 | static inline void clear_evtchn(int port) |
@@ -377,7 +377,7 @@ int bind_evtchn_to_irq(unsigned int evtchn) | |||
377 | irq = find_unbound_irq(); | 377 | irq = find_unbound_irq(); |
378 | 378 | ||
379 | set_irq_chip_and_handler_name(irq, &xen_dynamic_chip, | 379 | set_irq_chip_and_handler_name(irq, &xen_dynamic_chip, |
380 | handle_edge_irq, "event"); | 380 | handle_fasteoi_irq, "event"); |
381 | 381 | ||
382 | evtchn_to_irq[evtchn] = irq; | 382 | evtchn_to_irq[evtchn] = irq; |
383 | irq_info[irq] = mk_evtchn_info(evtchn); | 383 | irq_info[irq] = mk_evtchn_info(evtchn); |
@@ -435,6 +435,11 @@ static int bind_virq_to_irq(unsigned int virq, unsigned int cpu) | |||
435 | irq = per_cpu(virq_to_irq, cpu)[virq]; | 435 | irq = per_cpu(virq_to_irq, cpu)[virq]; |
436 | 436 | ||
437 | if (irq == -1) { | 437 | if (irq == -1) { |
438 | irq = find_unbound_irq(); | ||
439 | |||
440 | set_irq_chip_and_handler_name(irq, &xen_percpu_chip, | ||
441 | handle_percpu_irq, "virq"); | ||
442 | |||
438 | bind_virq.virq = virq; | 443 | bind_virq.virq = virq; |
439 | bind_virq.vcpu = cpu; | 444 | bind_virq.vcpu = cpu; |
440 | if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq, | 445 | if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq, |
@@ -442,11 +447,6 @@ static int bind_virq_to_irq(unsigned int virq, unsigned int cpu) | |||
442 | BUG(); | 447 | BUG(); |
443 | evtchn = bind_virq.port; | 448 | evtchn = bind_virq.port; |
444 | 449 | ||
445 | irq = find_unbound_irq(); | ||
446 | |||
447 | set_irq_chip_and_handler_name(irq, &xen_percpu_chip, | ||
448 | handle_percpu_irq, "virq"); | ||
449 | |||
450 | evtchn_to_irq[evtchn] = irq; | 450 | evtchn_to_irq[evtchn] = irq; |
451 | irq_info[irq] = mk_virq_info(evtchn, virq); | 451 | irq_info[irq] = mk_virq_info(evtchn, virq); |
452 | 452 | ||
@@ -578,41 +578,75 @@ irqreturn_t xen_debug_interrupt(int irq, void *dev_id) | |||
578 | { | 578 | { |
579 | struct shared_info *sh = HYPERVISOR_shared_info; | 579 | struct shared_info *sh = HYPERVISOR_shared_info; |
580 | int cpu = smp_processor_id(); | 580 | int cpu = smp_processor_id(); |
581 | unsigned long *cpu_evtchn = cpu_evtchn_mask(cpu); | ||
581 | int i; | 582 | int i; |
582 | unsigned long flags; | 583 | unsigned long flags; |
583 | static DEFINE_SPINLOCK(debug_lock); | 584 | static DEFINE_SPINLOCK(debug_lock); |
585 | struct vcpu_info *v; | ||
584 | 586 | ||
585 | spin_lock_irqsave(&debug_lock, flags); | 587 | spin_lock_irqsave(&debug_lock, flags); |
586 | 588 | ||
587 | printk("vcpu %d\n ", cpu); | 589 | printk("\nvcpu %d\n ", cpu); |
588 | 590 | ||
589 | for_each_online_cpu(i) { | 591 | for_each_online_cpu(i) { |
590 | struct vcpu_info *v = per_cpu(xen_vcpu, i); | 592 | int pending; |
591 | printk("%d: masked=%d pending=%d event_sel %08lx\n ", i, | 593 | v = per_cpu(xen_vcpu, i); |
592 | (get_irq_regs() && i == cpu) ? xen_irqs_disabled(get_irq_regs()) : v->evtchn_upcall_mask, | 594 | pending = (get_irq_regs() && i == cpu) |
593 | v->evtchn_upcall_pending, | 595 | ? xen_irqs_disabled(get_irq_regs()) |
594 | v->evtchn_pending_sel); | 596 | : v->evtchn_upcall_mask; |
597 | printk("%d: masked=%d pending=%d event_sel %0*lx\n ", i, | ||
598 | pending, v->evtchn_upcall_pending, | ||
599 | (int)(sizeof(v->evtchn_pending_sel)*2), | ||
600 | v->evtchn_pending_sel); | ||
601 | } | ||
602 | v = per_cpu(xen_vcpu, cpu); | ||
603 | |||
604 | printk("\npending:\n "); | ||
605 | for (i = ARRAY_SIZE(sh->evtchn_pending)-1; i >= 0; i--) | ||
606 | printk("%0*lx%s", (int)sizeof(sh->evtchn_pending[0])*2, | ||
607 | sh->evtchn_pending[i], | ||
608 | i % 8 == 0 ? "\n " : " "); | ||
609 | printk("\nglobal mask:\n "); | ||
610 | for (i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--) | ||
611 | printk("%0*lx%s", | ||
612 | (int)(sizeof(sh->evtchn_mask[0])*2), | ||
613 | sh->evtchn_mask[i], | ||
614 | i % 8 == 0 ? "\n " : " "); | ||
615 | |||
616 | printk("\nglobally unmasked:\n "); | ||
617 | for (i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--) | ||
618 | printk("%0*lx%s", (int)(sizeof(sh->evtchn_mask[0])*2), | ||
619 | sh->evtchn_pending[i] & ~sh->evtchn_mask[i], | ||
620 | i % 8 == 0 ? "\n " : " "); | ||
621 | |||
622 | printk("\nlocal cpu%d mask:\n ", cpu); | ||
623 | for (i = (NR_EVENT_CHANNELS/BITS_PER_LONG)-1; i >= 0; i--) | ||
624 | printk("%0*lx%s", (int)(sizeof(cpu_evtchn[0])*2), | ||
625 | cpu_evtchn[i], | ||
626 | i % 8 == 0 ? "\n " : " "); | ||
627 | |||
628 | printk("\nlocally unmasked:\n "); | ||
629 | for (i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--) { | ||
630 | unsigned long pending = sh->evtchn_pending[i] | ||
631 | & ~sh->evtchn_mask[i] | ||
632 | & cpu_evtchn[i]; | ||
633 | printk("%0*lx%s", (int)(sizeof(sh->evtchn_mask[0])*2), | ||
634 | pending, i % 8 == 0 ? "\n " : " "); | ||
595 | } | 635 | } |
596 | printk("pending:\n "); | ||
597 | for(i = ARRAY_SIZE(sh->evtchn_pending)-1; i >= 0; i--) | ||
598 | printk("%08lx%s", sh->evtchn_pending[i], | ||
599 | i % 8 == 0 ? "\n " : " "); | ||
600 | printk("\nmasks:\n "); | ||
601 | for(i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--) | ||
602 | printk("%08lx%s", sh->evtchn_mask[i], | ||
603 | i % 8 == 0 ? "\n " : " "); | ||
604 | |||
605 | printk("\nunmasked:\n "); | ||
606 | for(i = ARRAY_SIZE(sh->evtchn_mask)-1; i >= 0; i--) | ||
607 | printk("%08lx%s", sh->evtchn_pending[i] & ~sh->evtchn_mask[i], | ||
608 | i % 8 == 0 ? "\n " : " "); | ||
609 | 636 | ||
610 | printk("\npending list:\n"); | 637 | printk("\npending list:\n"); |
611 | for(i = 0; i < NR_EVENT_CHANNELS; i++) { | 638 | for (i = 0; i < NR_EVENT_CHANNELS; i++) { |
612 | if (sync_test_bit(i, sh->evtchn_pending)) { | 639 | if (sync_test_bit(i, sh->evtchn_pending)) { |
613 | printk(" %d: event %d -> irq %d\n", | 640 | int word_idx = i / BITS_PER_LONG; |
641 | printk(" %d: event %d -> irq %d%s%s%s\n", | ||
614 | cpu_from_evtchn(i), i, | 642 | cpu_from_evtchn(i), i, |
615 | evtchn_to_irq[i]); | 643 | evtchn_to_irq[i], |
644 | sync_test_bit(word_idx, &v->evtchn_pending_sel) | ||
645 | ? "" : " l2-clear", | ||
646 | !sync_test_bit(i, sh->evtchn_mask) | ||
647 | ? "" : " globally-masked", | ||
648 | sync_test_bit(i, cpu_evtchn) | ||
649 | ? "" : " locally-masked"); | ||
616 | } | 650 | } |
617 | } | 651 | } |
618 | 652 | ||
@@ -663,6 +697,9 @@ static void __xen_evtchn_do_upcall(void) | |||
663 | int irq = evtchn_to_irq[port]; | 697 | int irq = evtchn_to_irq[port]; |
664 | struct irq_desc *desc; | 698 | struct irq_desc *desc; |
665 | 699 | ||
700 | mask_evtchn(port); | ||
701 | clear_evtchn(port); | ||
702 | |||
666 | if (irq != -1) { | 703 | if (irq != -1) { |
667 | desc = irq_to_desc(irq); | 704 | desc = irq_to_desc(irq); |
668 | if (desc) | 705 | if (desc) |
@@ -800,10 +837,10 @@ static void ack_dynirq(unsigned int irq) | |||
800 | { | 837 | { |
801 | int evtchn = evtchn_from_irq(irq); | 838 | int evtchn = evtchn_from_irq(irq); |
802 | 839 | ||
803 | move_native_irq(irq); | 840 | move_masked_irq(irq); |
804 | 841 | ||
805 | if (VALID_EVTCHN(evtchn)) | 842 | if (VALID_EVTCHN(evtchn)) |
806 | clear_evtchn(evtchn); | 843 | unmask_evtchn(evtchn); |
807 | } | 844 | } |
808 | 845 | ||
809 | static int retrigger_dynirq(unsigned int irq) | 846 | static int retrigger_dynirq(unsigned int irq) |
@@ -959,7 +996,7 @@ static struct irq_chip xen_dynamic_chip __read_mostly = { | |||
959 | .mask = disable_dynirq, | 996 | .mask = disable_dynirq, |
960 | .unmask = enable_dynirq, | 997 | .unmask = enable_dynirq, |
961 | 998 | ||
962 | .ack = ack_dynirq, | 999 | .eoi = ack_dynirq, |
963 | .set_affinity = set_affinity_irq, | 1000 | .set_affinity = set_affinity_irq, |
964 | .retrigger = retrigger_dynirq, | 1001 | .retrigger = retrigger_dynirq, |
965 | }; | 1002 | }; |
diff --git a/drivers/xen/xenbus/xenbus_probe.c b/drivers/xen/xenbus/xenbus_probe.c index d409495876f1..132939f36020 100644 --- a/drivers/xen/xenbus/xenbus_probe.c +++ b/drivers/xen/xenbus/xenbus_probe.c | |||
@@ -64,9 +64,11 @@ | |||
64 | 64 | ||
65 | 65 | ||
66 | int xen_store_evtchn; | 66 | int xen_store_evtchn; |
67 | EXPORT_SYMBOL(xen_store_evtchn); | 67 | EXPORT_SYMBOL_GPL(xen_store_evtchn); |
68 | 68 | ||
69 | struct xenstore_domain_interface *xen_store_interface; | 69 | struct xenstore_domain_interface *xen_store_interface; |
70 | EXPORT_SYMBOL_GPL(xen_store_interface); | ||
71 | |||
70 | static unsigned long xen_store_mfn; | 72 | static unsigned long xen_store_mfn; |
71 | 73 | ||
72 | static BLOCKING_NOTIFIER_HEAD(xenstore_chain); | 74 | static BLOCKING_NOTIFIER_HEAD(xenstore_chain); |
diff --git a/drivers/xen/xenfs/Makefile b/drivers/xen/xenfs/Makefile index 25275c3bbdff..4fde9440fe1f 100644 --- a/drivers/xen/xenfs/Makefile +++ b/drivers/xen/xenfs/Makefile | |||
@@ -1,3 +1,4 @@ | |||
1 | obj-$(CONFIG_XENFS) += xenfs.o | 1 | obj-$(CONFIG_XENFS) += xenfs.o |
2 | 2 | ||
3 | xenfs-objs = super.o xenbus.o \ No newline at end of file | 3 | xenfs-y = super.o xenbus.o privcmd.o |
4 | xenfs-$(CONFIG_XEN_DOM0) += xenstored.o | ||
diff --git a/drivers/xen/xenfs/privcmd.c b/drivers/xen/xenfs/privcmd.c new file mode 100644 index 000000000000..f80be7f6eb95 --- /dev/null +++ b/drivers/xen/xenfs/privcmd.c | |||
@@ -0,0 +1,404 @@ | |||
1 | /****************************************************************************** | ||
2 | * privcmd.c | ||
3 | * | ||
4 | * Interface to privileged domain-0 commands. | ||
5 | * | ||
6 | * Copyright (c) 2002-2004, K A Fraser, B Dragovic | ||
7 | */ | ||
8 | |||
9 | #include <linux/kernel.h> | ||
10 | #include <linux/sched.h> | ||
11 | #include <linux/slab.h> | ||
12 | #include <linux/string.h> | ||
13 | #include <linux/errno.h> | ||
14 | #include <linux/mm.h> | ||
15 | #include <linux/mman.h> | ||
16 | #include <linux/uaccess.h> | ||
17 | #include <linux/swap.h> | ||
18 | #include <linux/smp_lock.h> | ||
19 | #include <linux/highmem.h> | ||
20 | #include <linux/pagemap.h> | ||
21 | #include <linux/seq_file.h> | ||
22 | |||
23 | #include <asm/pgalloc.h> | ||
24 | #include <asm/pgtable.h> | ||
25 | #include <asm/tlb.h> | ||
26 | #include <asm/xen/hypervisor.h> | ||
27 | #include <asm/xen/hypercall.h> | ||
28 | |||
29 | #include <xen/xen.h> | ||
30 | #include <xen/privcmd.h> | ||
31 | #include <xen/interface/xen.h> | ||
32 | #include <xen/features.h> | ||
33 | #include <xen/page.h> | ||
34 | #include <xen/xen-ops.h> | ||
35 | |||
36 | #ifndef HAVE_ARCH_PRIVCMD_MMAP | ||
37 | static int privcmd_enforce_singleshot_mapping(struct vm_area_struct *vma); | ||
38 | #endif | ||
39 | |||
40 | static long privcmd_ioctl_hypercall(void __user *udata) | ||
41 | { | ||
42 | struct privcmd_hypercall hypercall; | ||
43 | long ret; | ||
44 | |||
45 | if (copy_from_user(&hypercall, udata, sizeof(hypercall))) | ||
46 | return -EFAULT; | ||
47 | |||
48 | ret = privcmd_call(hypercall.op, | ||
49 | hypercall.arg[0], hypercall.arg[1], | ||
50 | hypercall.arg[2], hypercall.arg[3], | ||
51 | hypercall.arg[4]); | ||
52 | |||
53 | return ret; | ||
54 | } | ||
55 | |||
56 | static void free_page_list(struct list_head *pages) | ||
57 | { | ||
58 | struct page *p, *n; | ||
59 | |||
60 | list_for_each_entry_safe(p, n, pages, lru) | ||
61 | __free_page(p); | ||
62 | |||
63 | INIT_LIST_HEAD(pages); | ||
64 | } | ||
65 | |||
66 | /* | ||
67 | * Given an array of items in userspace, return a list of pages | ||
68 | * containing the data. If copying fails, either because of memory | ||
69 | * allocation failure or a problem reading user memory, return an | ||
70 | * error code; its up to the caller to dispose of any partial list. | ||
71 | */ | ||
72 | static int gather_array(struct list_head *pagelist, | ||
73 | unsigned nelem, size_t size, | ||
74 | void __user *data) | ||
75 | { | ||
76 | unsigned pageidx; | ||
77 | void *pagedata; | ||
78 | int ret; | ||
79 | |||
80 | if (size > PAGE_SIZE) | ||
81 | return 0; | ||
82 | |||
83 | pageidx = PAGE_SIZE; | ||
84 | pagedata = NULL; /* quiet, gcc */ | ||
85 | while (nelem--) { | ||
86 | if (pageidx > PAGE_SIZE-size) { | ||
87 | struct page *page = alloc_page(GFP_KERNEL); | ||
88 | |||
89 | ret = -ENOMEM; | ||
90 | if (page == NULL) | ||
91 | goto fail; | ||
92 | |||
93 | pagedata = page_address(page); | ||
94 | |||
95 | list_add_tail(&page->lru, pagelist); | ||
96 | pageidx = 0; | ||
97 | } | ||
98 | |||
99 | ret = -EFAULT; | ||
100 | if (copy_from_user(pagedata + pageidx, data, size)) | ||
101 | goto fail; | ||
102 | |||
103 | data += size; | ||
104 | pageidx += size; | ||
105 | } | ||
106 | |||
107 | ret = 0; | ||
108 | |||
109 | fail: | ||
110 | return ret; | ||
111 | } | ||
112 | |||
113 | /* | ||
114 | * Call function "fn" on each element of the array fragmented | ||
115 | * over a list of pages. | ||
116 | */ | ||
117 | static int traverse_pages(unsigned nelem, size_t size, | ||
118 | struct list_head *pos, | ||
119 | int (*fn)(void *data, void *state), | ||
120 | void *state) | ||
121 | { | ||
122 | void *pagedata; | ||
123 | unsigned pageidx; | ||
124 | int ret = 0; | ||
125 | |||
126 | BUG_ON(size > PAGE_SIZE); | ||
127 | |||
128 | pageidx = PAGE_SIZE; | ||
129 | pagedata = NULL; /* hush, gcc */ | ||
130 | |||
131 | while (nelem--) { | ||
132 | if (pageidx > PAGE_SIZE-size) { | ||
133 | struct page *page; | ||
134 | pos = pos->next; | ||
135 | page = list_entry(pos, struct page, lru); | ||
136 | pagedata = page_address(page); | ||
137 | pageidx = 0; | ||
138 | } | ||
139 | |||
140 | ret = (*fn)(pagedata + pageidx, state); | ||
141 | if (ret) | ||
142 | break; | ||
143 | pageidx += size; | ||
144 | } | ||
145 | |||
146 | return ret; | ||
147 | } | ||
148 | |||
149 | struct mmap_mfn_state { | ||
150 | unsigned long va; | ||
151 | struct vm_area_struct *vma; | ||
152 | domid_t domain; | ||
153 | }; | ||
154 | |||
155 | static int mmap_mfn_range(void *data, void *state) | ||
156 | { | ||
157 | struct privcmd_mmap_entry *msg = data; | ||
158 | struct mmap_mfn_state *st = state; | ||
159 | struct vm_area_struct *vma = st->vma; | ||
160 | int rc; | ||
161 | |||
162 | /* Do not allow range to wrap the address space. */ | ||
163 | if ((msg->npages > (LONG_MAX >> PAGE_SHIFT)) || | ||
164 | ((unsigned long)(msg->npages << PAGE_SHIFT) >= -st->va)) | ||
165 | return -EINVAL; | ||
166 | |||
167 | /* Range chunks must be contiguous in va space. */ | ||
168 | if ((msg->va != st->va) || | ||
169 | ((msg->va+(msg->npages<<PAGE_SHIFT)) > vma->vm_end)) | ||
170 | return -EINVAL; | ||
171 | |||
172 | rc = xen_remap_domain_mfn_range(vma, | ||
173 | msg->va & PAGE_MASK, | ||
174 | msg->mfn, msg->npages, | ||
175 | vma->vm_page_prot, | ||
176 | st->domain); | ||
177 | if (rc < 0) | ||
178 | return rc; | ||
179 | |||
180 | st->va += msg->npages << PAGE_SHIFT; | ||
181 | |||
182 | return 0; | ||
183 | } | ||
184 | |||
185 | static long privcmd_ioctl_mmap(void __user *udata) | ||
186 | { | ||
187 | struct privcmd_mmap mmapcmd; | ||
188 | struct mm_struct *mm = current->mm; | ||
189 | struct vm_area_struct *vma; | ||
190 | int rc; | ||
191 | LIST_HEAD(pagelist); | ||
192 | struct mmap_mfn_state state; | ||
193 | |||
194 | if (!xen_initial_domain()) | ||
195 | return -EPERM; | ||
196 | |||
197 | if (copy_from_user(&mmapcmd, udata, sizeof(mmapcmd))) | ||
198 | return -EFAULT; | ||
199 | |||
200 | rc = gather_array(&pagelist, | ||
201 | mmapcmd.num, sizeof(struct privcmd_mmap_entry), | ||
202 | mmapcmd.entry); | ||
203 | |||
204 | if (rc || list_empty(&pagelist)) | ||
205 | goto out; | ||
206 | |||
207 | down_write(&mm->mmap_sem); | ||
208 | |||
209 | { | ||
210 | struct page *page = list_first_entry(&pagelist, | ||
211 | struct page, lru); | ||
212 | struct privcmd_mmap_entry *msg = page_address(page); | ||
213 | |||
214 | vma = find_vma(mm, msg->va); | ||
215 | rc = -EINVAL; | ||
216 | |||
217 | if (!vma || (msg->va != vma->vm_start) || | ||
218 | !privcmd_enforce_singleshot_mapping(vma)) | ||
219 | goto out_up; | ||
220 | } | ||
221 | |||
222 | state.va = vma->vm_start; | ||
223 | state.vma = vma; | ||
224 | state.domain = mmapcmd.dom; | ||
225 | |||
226 | rc = traverse_pages(mmapcmd.num, sizeof(struct privcmd_mmap_entry), | ||
227 | &pagelist, | ||
228 | mmap_mfn_range, &state); | ||
229 | |||
230 | |||
231 | out_up: | ||
232 | up_write(&mm->mmap_sem); | ||
233 | |||
234 | out: | ||
235 | free_page_list(&pagelist); | ||
236 | |||
237 | return rc; | ||
238 | } | ||
239 | |||
240 | struct mmap_batch_state { | ||
241 | domid_t domain; | ||
242 | unsigned long va; | ||
243 | struct vm_area_struct *vma; | ||
244 | int err; | ||
245 | |||
246 | xen_pfn_t __user *user; | ||
247 | }; | ||
248 | |||
249 | static int mmap_batch_fn(void *data, void *state) | ||
250 | { | ||
251 | xen_pfn_t *mfnp = data; | ||
252 | struct mmap_batch_state *st = state; | ||
253 | |||
254 | if (xen_remap_domain_mfn_range(st->vma, st->va & PAGE_MASK, *mfnp, 1, | ||
255 | st->vma->vm_page_prot, st->domain) < 0) { | ||
256 | *mfnp |= 0xf0000000U; | ||
257 | st->err++; | ||
258 | } | ||
259 | st->va += PAGE_SIZE; | ||
260 | |||
261 | return 0; | ||
262 | } | ||
263 | |||
264 | static int mmap_return_errors(void *data, void *state) | ||
265 | { | ||
266 | xen_pfn_t *mfnp = data; | ||
267 | struct mmap_batch_state *st = state; | ||
268 | |||
269 | put_user(*mfnp, st->user++); | ||
270 | |||
271 | return 0; | ||
272 | } | ||
273 | |||
274 | static struct vm_operations_struct privcmd_vm_ops; | ||
275 | |||
276 | static long privcmd_ioctl_mmap_batch(void __user *udata) | ||
277 | { | ||
278 | int ret; | ||
279 | struct privcmd_mmapbatch m; | ||
280 | struct mm_struct *mm = current->mm; | ||
281 | struct vm_area_struct *vma; | ||
282 | unsigned long nr_pages; | ||
283 | LIST_HEAD(pagelist); | ||
284 | struct mmap_batch_state state; | ||
285 | |||
286 | if (!xen_initial_domain()) | ||
287 | return -EPERM; | ||
288 | |||
289 | if (copy_from_user(&m, udata, sizeof(m))) | ||
290 | return -EFAULT; | ||
291 | |||
292 | nr_pages = m.num; | ||
293 | if ((m.num <= 0) || (nr_pages > (LONG_MAX >> PAGE_SHIFT))) | ||
294 | return -EINVAL; | ||
295 | |||
296 | ret = gather_array(&pagelist, m.num, sizeof(xen_pfn_t), | ||
297 | m.arr); | ||
298 | |||
299 | if (ret || list_empty(&pagelist)) | ||
300 | goto out; | ||
301 | |||
302 | down_write(&mm->mmap_sem); | ||
303 | |||
304 | vma = find_vma(mm, m.addr); | ||
305 | ret = -EINVAL; | ||
306 | if (!vma || | ||
307 | vma->vm_ops != &privcmd_vm_ops || | ||
308 | (m.addr != vma->vm_start) || | ||
309 | ((m.addr + (nr_pages << PAGE_SHIFT)) != vma->vm_end) || | ||
310 | !privcmd_enforce_singleshot_mapping(vma)) { | ||
311 | up_write(&mm->mmap_sem); | ||
312 | goto out; | ||
313 | } | ||
314 | |||
315 | state.domain = m.dom; | ||
316 | state.vma = vma; | ||
317 | state.va = m.addr; | ||
318 | state.err = 0; | ||
319 | |||
320 | ret = traverse_pages(m.num, sizeof(xen_pfn_t), | ||
321 | &pagelist, mmap_batch_fn, &state); | ||
322 | |||
323 | up_write(&mm->mmap_sem); | ||
324 | |||
325 | if (state.err > 0) { | ||
326 | ret = 0; | ||
327 | |||
328 | state.user = m.arr; | ||
329 | traverse_pages(m.num, sizeof(xen_pfn_t), | ||
330 | &pagelist, | ||
331 | mmap_return_errors, &state); | ||
332 | } | ||
333 | |||
334 | out: | ||
335 | free_page_list(&pagelist); | ||
336 | |||
337 | return ret; | ||
338 | } | ||
339 | |||
340 | static long privcmd_ioctl(struct file *file, | ||
341 | unsigned int cmd, unsigned long data) | ||
342 | { | ||
343 | int ret = -ENOSYS; | ||
344 | void __user *udata = (void __user *) data; | ||
345 | |||
346 | switch (cmd) { | ||
347 | case IOCTL_PRIVCMD_HYPERCALL: | ||
348 | ret = privcmd_ioctl_hypercall(udata); | ||
349 | break; | ||
350 | |||
351 | case IOCTL_PRIVCMD_MMAP: | ||
352 | ret = privcmd_ioctl_mmap(udata); | ||
353 | break; | ||
354 | |||
355 | case IOCTL_PRIVCMD_MMAPBATCH: | ||
356 | ret = privcmd_ioctl_mmap_batch(udata); | ||
357 | break; | ||
358 | |||
359 | default: | ||
360 | ret = -EINVAL; | ||
361 | break; | ||
362 | } | ||
363 | |||
364 | return ret; | ||
365 | } | ||
366 | |||
367 | #ifndef HAVE_ARCH_PRIVCMD_MMAP | ||
368 | static int privcmd_fault(struct vm_area_struct *vma, struct vm_fault *vmf) | ||
369 | { | ||
370 | printk(KERN_DEBUG "privcmd_fault: vma=%p %lx-%lx, pgoff=%lx, uv=%p\n", | ||
371 | vma, vma->vm_start, vma->vm_end, | ||
372 | vmf->pgoff, vmf->virtual_address); | ||
373 | |||
374 | return VM_FAULT_SIGBUS; | ||
375 | } | ||
376 | |||
377 | static struct vm_operations_struct privcmd_vm_ops = { | ||
378 | .fault = privcmd_fault | ||
379 | }; | ||
380 | |||
381 | static int privcmd_mmap(struct file *file, struct vm_area_struct *vma) | ||
382 | { | ||
383 | /* Unsupported for auto-translate guests. */ | ||
384 | if (xen_feature(XENFEAT_auto_translated_physmap)) | ||
385 | return -ENOSYS; | ||
386 | |||
387 | /* DONTCOPY is essential for Xen as copy_page_range is broken. */ | ||
388 | vma->vm_flags |= VM_RESERVED | VM_IO | VM_DONTCOPY; | ||
389 | vma->vm_ops = &privcmd_vm_ops; | ||
390 | vma->vm_private_data = NULL; | ||
391 | |||
392 | return 0; | ||
393 | } | ||
394 | |||
395 | static int privcmd_enforce_singleshot_mapping(struct vm_area_struct *vma) | ||
396 | { | ||
397 | return (xchg(&vma->vm_private_data, (void *)1) == NULL); | ||
398 | } | ||
399 | #endif | ||
400 | |||
401 | const struct file_operations privcmd_file_ops = { | ||
402 | .unlocked_ioctl = privcmd_ioctl, | ||
403 | .mmap = privcmd_mmap, | ||
404 | }; | ||
diff --git a/drivers/xen/xenfs/super.c b/drivers/xen/xenfs/super.c index bd96340063c1..d6662b789b6b 100644 --- a/drivers/xen/xenfs/super.c +++ b/drivers/xen/xenfs/super.c | |||
@@ -12,6 +12,8 @@ | |||
12 | #include <linux/module.h> | 12 | #include <linux/module.h> |
13 | #include <linux/fs.h> | 13 | #include <linux/fs.h> |
14 | #include <linux/magic.h> | 14 | #include <linux/magic.h> |
15 | #include <linux/mm.h> | ||
16 | #include <linux/backing-dev.h> | ||
15 | 17 | ||
16 | #include <xen/xen.h> | 18 | #include <xen/xen.h> |
17 | 19 | ||
@@ -22,6 +24,62 @@ | |||
22 | MODULE_DESCRIPTION("Xen filesystem"); | 24 | MODULE_DESCRIPTION("Xen filesystem"); |
23 | MODULE_LICENSE("GPL"); | 25 | MODULE_LICENSE("GPL"); |
24 | 26 | ||
27 | static int xenfs_set_page_dirty(struct page *page) | ||
28 | { | ||
29 | return !TestSetPageDirty(page); | ||
30 | } | ||
31 | |||
32 | static const struct address_space_operations xenfs_aops = { | ||
33 | .set_page_dirty = xenfs_set_page_dirty, | ||
34 | }; | ||
35 | |||
36 | static struct backing_dev_info xenfs_backing_dev_info = { | ||
37 | .ra_pages = 0, /* No readahead */ | ||
38 | .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK, | ||
39 | }; | ||
40 | |||
41 | static struct inode *xenfs_make_inode(struct super_block *sb, int mode) | ||
42 | { | ||
43 | struct inode *ret = new_inode(sb); | ||
44 | |||
45 | if (ret) { | ||
46 | ret->i_mode = mode; | ||
47 | ret->i_mapping->a_ops = &xenfs_aops; | ||
48 | ret->i_mapping->backing_dev_info = &xenfs_backing_dev_info; | ||
49 | ret->i_uid = ret->i_gid = 0; | ||
50 | ret->i_blocks = 0; | ||
51 | ret->i_atime = ret->i_mtime = ret->i_ctime = CURRENT_TIME; | ||
52 | } | ||
53 | return ret; | ||
54 | } | ||
55 | |||
56 | static struct dentry *xenfs_create_file(struct super_block *sb, | ||
57 | struct dentry *parent, | ||
58 | const char *name, | ||
59 | const struct file_operations *fops, | ||
60 | void *data, | ||
61 | int mode) | ||
62 | { | ||
63 | struct dentry *dentry; | ||
64 | struct inode *inode; | ||
65 | |||
66 | dentry = d_alloc_name(parent, name); | ||
67 | if (!dentry) | ||
68 | return NULL; | ||
69 | |||
70 | inode = xenfs_make_inode(sb, S_IFREG | mode); | ||
71 | if (!inode) { | ||
72 | dput(dentry); | ||
73 | return NULL; | ||
74 | } | ||
75 | |||
76 | inode->i_fop = fops; | ||
77 | inode->i_private = data; | ||
78 | |||
79 | d_add(dentry, inode); | ||
80 | return dentry; | ||
81 | } | ||
82 | |||
25 | static ssize_t capabilities_read(struct file *file, char __user *buf, | 83 | static ssize_t capabilities_read(struct file *file, char __user *buf, |
26 | size_t size, loff_t *off) | 84 | size_t size, loff_t *off) |
27 | { | 85 | { |
@@ -44,10 +102,23 @@ static int xenfs_fill_super(struct super_block *sb, void *data, int silent) | |||
44 | [1] = {}, | 102 | [1] = {}, |
45 | { "xenbus", &xenbus_file_ops, S_IRUSR|S_IWUSR }, | 103 | { "xenbus", &xenbus_file_ops, S_IRUSR|S_IWUSR }, |
46 | { "capabilities", &capabilities_file_ops, S_IRUGO }, | 104 | { "capabilities", &capabilities_file_ops, S_IRUGO }, |
105 | { "privcmd", &privcmd_file_ops, S_IRUSR|S_IWUSR }, | ||
47 | {""}, | 106 | {""}, |
48 | }; | 107 | }; |
108 | int rc; | ||
49 | 109 | ||
50 | return simple_fill_super(sb, XENFS_SUPER_MAGIC, xenfs_files); | 110 | rc = simple_fill_super(sb, XENFS_SUPER_MAGIC, xenfs_files); |
111 | if (rc < 0) | ||
112 | return rc; | ||
113 | |||
114 | if (xen_initial_domain()) { | ||
115 | xenfs_create_file(sb, sb->s_root, "xsd_kva", | ||
116 | &xsd_kva_file_ops, NULL, S_IRUSR|S_IWUSR); | ||
117 | xenfs_create_file(sb, sb->s_root, "xsd_port", | ||
118 | &xsd_port_file_ops, NULL, S_IRUSR|S_IWUSR); | ||
119 | } | ||
120 | |||
121 | return rc; | ||
51 | } | 122 | } |
52 | 123 | ||
53 | static int xenfs_get_sb(struct file_system_type *fs_type, | 124 | static int xenfs_get_sb(struct file_system_type *fs_type, |
@@ -66,11 +137,25 @@ static struct file_system_type xenfs_type = { | |||
66 | 137 | ||
67 | static int __init xenfs_init(void) | 138 | static int __init xenfs_init(void) |
68 | { | 139 | { |
69 | if (xen_domain()) | 140 | int err; |
70 | return register_filesystem(&xenfs_type); | 141 | if (!xen_domain()) { |
142 | printk(KERN_INFO "xenfs: not registering filesystem on non-xen platform\n"); | ||
143 | return 0; | ||
144 | } | ||
145 | |||
146 | err = register_filesystem(&xenfs_type); | ||
147 | if (err) { | ||
148 | printk(KERN_ERR "xenfs: Unable to register filesystem!\n"); | ||
149 | goto out; | ||
150 | } | ||
151 | |||
152 | err = bdi_init(&xenfs_backing_dev_info); | ||
153 | if (err) | ||
154 | unregister_filesystem(&xenfs_type); | ||
155 | |||
156 | out: | ||
71 | 157 | ||
72 | printk(KERN_INFO "XENFS: not registering filesystem on non-xen platform\n"); | 158 | return err; |
73 | return 0; | ||
74 | } | 159 | } |
75 | 160 | ||
76 | static void __exit xenfs_exit(void) | 161 | static void __exit xenfs_exit(void) |
diff --git a/drivers/xen/xenfs/xenfs.h b/drivers/xen/xenfs/xenfs.h index 51f08b2d0bf1..b68aa6200003 100644 --- a/drivers/xen/xenfs/xenfs.h +++ b/drivers/xen/xenfs/xenfs.h | |||
@@ -2,5 +2,8 @@ | |||
2 | #define _XENFS_XENBUS_H | 2 | #define _XENFS_XENBUS_H |
3 | 3 | ||
4 | extern const struct file_operations xenbus_file_ops; | 4 | extern const struct file_operations xenbus_file_ops; |
5 | extern const struct file_operations privcmd_file_ops; | ||
6 | extern const struct file_operations xsd_kva_file_ops; | ||
7 | extern const struct file_operations xsd_port_file_ops; | ||
5 | 8 | ||
6 | #endif /* _XENFS_XENBUS_H */ | 9 | #endif /* _XENFS_XENBUS_H */ |
diff --git a/drivers/xen/xenfs/xenstored.c b/drivers/xen/xenfs/xenstored.c new file mode 100644 index 000000000000..fef20dbc6a5c --- /dev/null +++ b/drivers/xen/xenfs/xenstored.c | |||
@@ -0,0 +1,68 @@ | |||
1 | #include <linux/slab.h> | ||
2 | #include <linux/types.h> | ||
3 | #include <linux/mm.h> | ||
4 | #include <linux/fs.h> | ||
5 | |||
6 | #include <xen/page.h> | ||
7 | |||
8 | #include "xenfs.h" | ||
9 | #include "../xenbus/xenbus_comms.h" | ||
10 | |||
11 | static ssize_t xsd_read(struct file *file, char __user *buf, | ||
12 | size_t size, loff_t *off) | ||
13 | { | ||
14 | const char *str = (const char *)file->private_data; | ||
15 | return simple_read_from_buffer(buf, size, off, str, strlen(str)); | ||
16 | } | ||
17 | |||
18 | static int xsd_release(struct inode *inode, struct file *file) | ||
19 | { | ||
20 | kfree(file->private_data); | ||
21 | return 0; | ||
22 | } | ||
23 | |||
24 | static int xsd_kva_open(struct inode *inode, struct file *file) | ||
25 | { | ||
26 | file->private_data = (void *)kasprintf(GFP_KERNEL, "0x%p", | ||
27 | xen_store_interface); | ||
28 | if (!file->private_data) | ||
29 | return -ENOMEM; | ||
30 | return 0; | ||
31 | } | ||
32 | |||
33 | static int xsd_kva_mmap(struct file *file, struct vm_area_struct *vma) | ||
34 | { | ||
35 | size_t size = vma->vm_end - vma->vm_start; | ||
36 | |||
37 | if ((size > PAGE_SIZE) || (vma->vm_pgoff != 0)) | ||
38 | return -EINVAL; | ||
39 | |||
40 | if (remap_pfn_range(vma, vma->vm_start, | ||
41 | virt_to_pfn(xen_store_interface), | ||
42 | size, vma->vm_page_prot)) | ||
43 | return -EAGAIN; | ||
44 | |||
45 | return 0; | ||
46 | } | ||
47 | |||
48 | const struct file_operations xsd_kva_file_ops = { | ||
49 | .open = xsd_kva_open, | ||
50 | .mmap = xsd_kva_mmap, | ||
51 | .read = xsd_read, | ||
52 | .release = xsd_release, | ||
53 | }; | ||
54 | |||
55 | static int xsd_port_open(struct inode *inode, struct file *file) | ||
56 | { | ||
57 | file->private_data = (void *)kasprintf(GFP_KERNEL, "%d", | ||
58 | xen_store_evtchn); | ||
59 | if (!file->private_data) | ||
60 | return -ENOMEM; | ||
61 | return 0; | ||
62 | } | ||
63 | |||
64 | const struct file_operations xsd_port_file_ops = { | ||
65 | .open = xsd_port_open, | ||
66 | .read = xsd_read, | ||
67 | .release = xsd_release, | ||
68 | }; | ||