diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2012-10-23 21:07:51 -0400 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2012-10-23 21:07:51 -0400 |
| commit | e17b1315832b97700502200f0a3dde3ac0fbc839 (patch) | |
| tree | e5a9bf48b596f327f303539b264584c7fd5ebca4 | |
| parent | 684baeb1d78ac478a8e5f00aa9ab0ce0837f2a91 (diff) | |
| parent | ef8ff74ed8dd9d4b3ba8cb9f2fc927a27c697a8b (diff) | |
Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull perf fixes from Ingo Molnar:
"Most of these are uprobes race fixes from Oleg, and their preparatory
cleanups. (It's larger than what I'd normally send for an -rc kernel,
but they looked significant enough to not delay them.)
There's also an oprofile fix and an uncore PMU fix."
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (22 commits)
perf/x86: Disable uncore on virtualized CPUs
oprofile, x86: Fix wrapping bug in op_x86_get_ctrl()
ring-buffer: Check for uninitialized cpu buffer before resizing
uprobes: Fix the racy uprobe->flags manipulation
uprobes: Fix prepare_uprobe() race with itself
uprobes: Introduce prepare_uprobe()
uprobes: Fix handle_swbp() vs unregister() + register() race
uprobes: Do not delete uprobe if uprobe_unregister() fails
uprobes: Don't return success if alloc_uprobe() fails
uprobes/x86: Only rep+nop can be emulated correctly
uprobes: Simplify is_swbp_at_addr(), remove stale comments
uprobes: Kill set_orig_insn()->is_swbp_at_addr()
uprobes: Introduce copy_opcode(), kill read_opcode()
uprobes: Kill set_swbp()->is_swbp_at_addr()
uprobes: Restrict valid_vma(false) to skip VM_SHARED vmas
uprobes: Change valid_vma() to demand VM_MAYEXEC rather than VM_EXEC
uprobes: Change write_opcode() to use FOLL_FORCE
uprobes: Move clear_thread_flag(TIF_UPROBE) to uprobe_notify_resume()
uprobes: Kill UTASK_BP_HIT state
uprobes: Fix UPROBE_SKIP_SSTEP checks in handle_swbp()
...
| -rw-r--r-- | arch/x86/kernel/cpu/perf_event_intel_uncore.c | 3 | ||||
| -rw-r--r-- | arch/x86/kernel/signal.c | 4 | ||||
| -rw-r--r-- | arch/x86/kernel/uprobes.c | 16 | ||||
| -rw-r--r-- | arch/x86/oprofile/nmi_int.c | 2 | ||||
| -rw-r--r-- | include/linux/uprobes.h | 11 | ||||
| -rw-r--r-- | kernel/events/uprobes.c | 345 | ||||
| -rw-r--r-- | kernel/trace/ring_buffer.c | 4 |
7 files changed, 173 insertions, 212 deletions
diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.c b/arch/x86/kernel/cpu/perf_event_intel_uncore.c index 99d96a4978b5..5df8d32ba91e 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_uncore.c +++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.c | |||
| @@ -2926,6 +2926,9 @@ static int __init intel_uncore_init(void) | |||
| 2926 | if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) | 2926 | if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) |
| 2927 | return -ENODEV; | 2927 | return -ENODEV; |
| 2928 | 2928 | ||
| 2929 | if (cpu_has_hypervisor) | ||
| 2930 | return -ENODEV; | ||
| 2931 | |||
| 2929 | ret = uncore_pci_init(); | 2932 | ret = uncore_pci_init(); |
| 2930 | if (ret) | 2933 | if (ret) |
| 2931 | goto fail; | 2934 | goto fail; |
diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c index 29ad351804e9..70b27ee6118e 100644 --- a/arch/x86/kernel/signal.c +++ b/arch/x86/kernel/signal.c | |||
| @@ -824,10 +824,8 @@ do_notify_resume(struct pt_regs *regs, void *unused, __u32 thread_info_flags) | |||
| 824 | mce_notify_process(); | 824 | mce_notify_process(); |
| 825 | #endif /* CONFIG_X86_64 && CONFIG_X86_MCE */ | 825 | #endif /* CONFIG_X86_64 && CONFIG_X86_MCE */ |
| 826 | 826 | ||
| 827 | if (thread_info_flags & _TIF_UPROBE) { | 827 | if (thread_info_flags & _TIF_UPROBE) |
| 828 | clear_thread_flag(TIF_UPROBE); | ||
| 829 | uprobe_notify_resume(regs); | 828 | uprobe_notify_resume(regs); |
| 830 | } | ||
| 831 | 829 | ||
| 832 | /* deal with pending signal delivery */ | 830 | /* deal with pending signal delivery */ |
| 833 | if (thread_info_flags & _TIF_SIGPENDING) | 831 | if (thread_info_flags & _TIF_SIGPENDING) |
diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c index 9538f00827a9..aafa5557b396 100644 --- a/arch/x86/kernel/uprobes.c +++ b/arch/x86/kernel/uprobes.c | |||
| @@ -651,31 +651,19 @@ void arch_uprobe_abort_xol(struct arch_uprobe *auprobe, struct pt_regs *regs) | |||
| 651 | 651 | ||
| 652 | /* | 652 | /* |
| 653 | * Skip these instructions as per the currently known x86 ISA. | 653 | * Skip these instructions as per the currently known x86 ISA. |
| 654 | * 0x66* { 0x90 | 0x0f 0x1f | 0x0f 0x19 | 0x87 0xc0 } | 654 | * rep=0x66*; nop=0x90 |
| 655 | */ | 655 | */ |
| 656 | static bool __skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs) | 656 | static bool __skip_sstep(struct arch_uprobe *auprobe, struct pt_regs *regs) |
| 657 | { | 657 | { |
| 658 | int i; | 658 | int i; |
| 659 | 659 | ||
| 660 | for (i = 0; i < MAX_UINSN_BYTES; i++) { | 660 | for (i = 0; i < MAX_UINSN_BYTES; i++) { |
| 661 | if ((auprobe->insn[i] == 0x66)) | 661 | if (auprobe->insn[i] == 0x66) |
| 662 | continue; | 662 | continue; |
| 663 | 663 | ||
| 664 | if (auprobe->insn[i] == 0x90) | 664 | if (auprobe->insn[i] == 0x90) |
| 665 | return true; | 665 | return true; |
| 666 | 666 | ||
| 667 | if (i == (MAX_UINSN_BYTES - 1)) | ||
| 668 | break; | ||
| 669 | |||
| 670 | if ((auprobe->insn[i] == 0x0f) && (auprobe->insn[i+1] == 0x1f)) | ||
| 671 | return true; | ||
| 672 | |||
| 673 | if ((auprobe->insn[i] == 0x0f) && (auprobe->insn[i+1] == 0x19)) | ||
| 674 | return true; | ||
| 675 | |||
| 676 | if ((auprobe->insn[i] == 0x87) && (auprobe->insn[i+1] == 0xc0)) | ||
| 677 | return true; | ||
| 678 | |||
| 679 | break; | 667 | break; |
| 680 | } | 668 | } |
| 681 | return false; | 669 | return false; |
diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c index 26b8a8514ee5..48768df2471a 100644 --- a/arch/x86/oprofile/nmi_int.c +++ b/arch/x86/oprofile/nmi_int.c | |||
| @@ -55,7 +55,7 @@ u64 op_x86_get_ctrl(struct op_x86_model_spec const *model, | |||
| 55 | val |= counter_config->extra; | 55 | val |= counter_config->extra; |
| 56 | event &= model->event_mask ? model->event_mask : 0xFF; | 56 | event &= model->event_mask ? model->event_mask : 0xFF; |
| 57 | val |= event & 0xFF; | 57 | val |= event & 0xFF; |
| 58 | val |= (event & 0x0F00) << 24; | 58 | val |= (u64)(event & 0x0F00) << 24; |
| 59 | 59 | ||
| 60 | return val; | 60 | return val; |
| 61 | } | 61 | } |
diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h index e6f0331e3d45..24594571c5a3 100644 --- a/include/linux/uprobes.h +++ b/include/linux/uprobes.h | |||
| @@ -35,16 +35,6 @@ struct inode; | |||
| 35 | # include <asm/uprobes.h> | 35 | # include <asm/uprobes.h> |
| 36 | #endif | 36 | #endif |
| 37 | 37 | ||
| 38 | /* flags that denote/change uprobes behaviour */ | ||
| 39 | |||
| 40 | /* Have a copy of original instruction */ | ||
| 41 | #define UPROBE_COPY_INSN 0x1 | ||
| 42 | |||
| 43 | /* Dont run handlers when first register/ last unregister in progress*/ | ||
| 44 | #define UPROBE_RUN_HANDLER 0x2 | ||
| 45 | /* Can skip singlestep */ | ||
| 46 | #define UPROBE_SKIP_SSTEP 0x4 | ||
| 47 | |||
| 48 | struct uprobe_consumer { | 38 | struct uprobe_consumer { |
| 49 | int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs); | 39 | int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs); |
| 50 | /* | 40 | /* |
| @@ -59,7 +49,6 @@ struct uprobe_consumer { | |||
| 59 | #ifdef CONFIG_UPROBES | 49 | #ifdef CONFIG_UPROBES |
| 60 | enum uprobe_task_state { | 50 | enum uprobe_task_state { |
| 61 | UTASK_RUNNING, | 51 | UTASK_RUNNING, |
| 62 | UTASK_BP_HIT, | ||
| 63 | UTASK_SSTEP, | 52 | UTASK_SSTEP, |
| 64 | UTASK_SSTEP_ACK, | 53 | UTASK_SSTEP_ACK, |
| 65 | UTASK_SSTEP_TRAPPED, | 54 | UTASK_SSTEP_TRAPPED, |
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c index 98256bc71ee1..5cc4e7e42e68 100644 --- a/kernel/events/uprobes.c +++ b/kernel/events/uprobes.c | |||
| @@ -78,15 +78,23 @@ static struct mutex uprobes_mmap_mutex[UPROBES_HASH_SZ]; | |||
| 78 | */ | 78 | */ |
| 79 | static atomic_t uprobe_events = ATOMIC_INIT(0); | 79 | static atomic_t uprobe_events = ATOMIC_INIT(0); |
| 80 | 80 | ||
| 81 | /* Have a copy of original instruction */ | ||
| 82 | #define UPROBE_COPY_INSN 0 | ||
| 83 | /* Dont run handlers when first register/ last unregister in progress*/ | ||
| 84 | #define UPROBE_RUN_HANDLER 1 | ||
| 85 | /* Can skip singlestep */ | ||
| 86 | #define UPROBE_SKIP_SSTEP 2 | ||
| 87 | |||
| 81 | struct uprobe { | 88 | struct uprobe { |
| 82 | struct rb_node rb_node; /* node in the rb tree */ | 89 | struct rb_node rb_node; /* node in the rb tree */ |
| 83 | atomic_t ref; | 90 | atomic_t ref; |
| 84 | struct rw_semaphore consumer_rwsem; | 91 | struct rw_semaphore consumer_rwsem; |
| 92 | struct mutex copy_mutex; /* TODO: kill me and UPROBE_COPY_INSN */ | ||
| 85 | struct list_head pending_list; | 93 | struct list_head pending_list; |
| 86 | struct uprobe_consumer *consumers; | 94 | struct uprobe_consumer *consumers; |
| 87 | struct inode *inode; /* Also hold a ref to inode */ | 95 | struct inode *inode; /* Also hold a ref to inode */ |
| 88 | loff_t offset; | 96 | loff_t offset; |
| 89 | int flags; | 97 | unsigned long flags; |
| 90 | struct arch_uprobe arch; | 98 | struct arch_uprobe arch; |
| 91 | }; | 99 | }; |
| 92 | 100 | ||
| @@ -100,17 +108,12 @@ struct uprobe { | |||
| 100 | */ | 108 | */ |
| 101 | static bool valid_vma(struct vm_area_struct *vma, bool is_register) | 109 | static bool valid_vma(struct vm_area_struct *vma, bool is_register) |
| 102 | { | 110 | { |
| 103 | if (!vma->vm_file) | 111 | vm_flags_t flags = VM_HUGETLB | VM_MAYEXEC | VM_SHARED; |
| 104 | return false; | ||
| 105 | |||
| 106 | if (!is_register) | ||
| 107 | return true; | ||
| 108 | 112 | ||
| 109 | if ((vma->vm_flags & (VM_HUGETLB|VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)) | 113 | if (is_register) |
| 110 | == (VM_READ|VM_EXEC)) | 114 | flags |= VM_WRITE; |
| 111 | return true; | ||
| 112 | 115 | ||
| 113 | return false; | 116 | return vma->vm_file && (vma->vm_flags & flags) == VM_MAYEXEC; |
| 114 | } | 117 | } |
| 115 | 118 | ||
| 116 | static unsigned long offset_to_vaddr(struct vm_area_struct *vma, loff_t offset) | 119 | static unsigned long offset_to_vaddr(struct vm_area_struct *vma, loff_t offset) |
| @@ -193,19 +196,44 @@ bool __weak is_swbp_insn(uprobe_opcode_t *insn) | |||
| 193 | return *insn == UPROBE_SWBP_INSN; | 196 | return *insn == UPROBE_SWBP_INSN; |
| 194 | } | 197 | } |
| 195 | 198 | ||
| 199 | static void copy_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t *opcode) | ||
| 200 | { | ||
| 201 | void *kaddr = kmap_atomic(page); | ||
| 202 | memcpy(opcode, kaddr + (vaddr & ~PAGE_MASK), UPROBE_SWBP_INSN_SIZE); | ||
| 203 | kunmap_atomic(kaddr); | ||
| 204 | } | ||
| 205 | |||
| 206 | static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t *new_opcode) | ||
| 207 | { | ||
| 208 | uprobe_opcode_t old_opcode; | ||
| 209 | bool is_swbp; | ||
| 210 | |||
| 211 | copy_opcode(page, vaddr, &old_opcode); | ||
| 212 | is_swbp = is_swbp_insn(&old_opcode); | ||
| 213 | |||
| 214 | if (is_swbp_insn(new_opcode)) { | ||
| 215 | if (is_swbp) /* register: already installed? */ | ||
| 216 | return 0; | ||
| 217 | } else { | ||
| 218 | if (!is_swbp) /* unregister: was it changed by us? */ | ||
| 219 | return 0; | ||
| 220 | } | ||
| 221 | |||
| 222 | return 1; | ||
| 223 | } | ||
| 224 | |||
| 196 | /* | 225 | /* |
| 197 | * NOTE: | 226 | * NOTE: |
| 198 | * Expect the breakpoint instruction to be the smallest size instruction for | 227 | * Expect the breakpoint instruction to be the smallest size instruction for |
| 199 | * the architecture. If an arch has variable length instruction and the | 228 | * the architecture. If an arch has variable length instruction and the |
| 200 | * breakpoint instruction is not of the smallest length instruction | 229 | * breakpoint instruction is not of the smallest length instruction |
| 201 | * supported by that architecture then we need to modify read_opcode / | 230 | * supported by that architecture then we need to modify is_swbp_at_addr and |
| 202 | * write_opcode accordingly. This would never be a problem for archs that | 231 | * write_opcode accordingly. This would never be a problem for archs that |
| 203 | * have fixed length instructions. | 232 | * have fixed length instructions. |
| 204 | */ | 233 | */ |
| 205 | 234 | ||
| 206 | /* | 235 | /* |
| 207 | * write_opcode - write the opcode at a given virtual address. | 236 | * write_opcode - write the opcode at a given virtual address. |
| 208 | * @auprobe: arch breakpointing information. | ||
| 209 | * @mm: the probed process address space. | 237 | * @mm: the probed process address space. |
| 210 | * @vaddr: the virtual address to store the opcode. | 238 | * @vaddr: the virtual address to store the opcode. |
| 211 | * @opcode: opcode to be written at @vaddr. | 239 | * @opcode: opcode to be written at @vaddr. |
| @@ -216,8 +244,8 @@ bool __weak is_swbp_insn(uprobe_opcode_t *insn) | |||
| 216 | * For mm @mm, write the opcode at @vaddr. | 244 | * For mm @mm, write the opcode at @vaddr. |
| 217 | * Return 0 (success) or a negative errno. | 245 | * Return 0 (success) or a negative errno. |
| 218 | */ | 246 | */ |
| 219 | static int write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm, | 247 | static int write_opcode(struct mm_struct *mm, unsigned long vaddr, |
| 220 | unsigned long vaddr, uprobe_opcode_t opcode) | 248 | uprobe_opcode_t opcode) |
| 221 | { | 249 | { |
| 222 | struct page *old_page, *new_page; | 250 | struct page *old_page, *new_page; |
| 223 | void *vaddr_old, *vaddr_new; | 251 | void *vaddr_old, *vaddr_new; |
| @@ -226,10 +254,14 @@ static int write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm, | |||
| 226 | 254 | ||
| 227 | retry: | 255 | retry: |
| 228 | /* Read the page with vaddr into memory */ | 256 | /* Read the page with vaddr into memory */ |
| 229 | ret = get_user_pages(NULL, mm, vaddr, 1, 0, 0, &old_page, &vma); | 257 | ret = get_user_pages(NULL, mm, vaddr, 1, 0, 1, &old_page, &vma); |
| 230 | if (ret <= 0) | 258 | if (ret <= 0) |
| 231 | return ret; | 259 | return ret; |
| 232 | 260 | ||
| 261 | ret = verify_opcode(old_page, vaddr, &opcode); | ||
| 262 | if (ret <= 0) | ||
| 263 | goto put_old; | ||
| 264 | |||
| 233 | ret = -ENOMEM; | 265 | ret = -ENOMEM; |
| 234 | new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vaddr); | 266 | new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vaddr); |
| 235 | if (!new_page) | 267 | if (!new_page) |
| @@ -264,63 +296,6 @@ put_old: | |||
| 264 | } | 296 | } |
| 265 | 297 | ||
| 266 | /** | 298 | /** |
| 267 | * read_opcode - read the opcode at a given virtual address. | ||
| 268 | * @mm: the probed process address space. | ||
| 269 | * @vaddr: the virtual address to read the opcode. | ||
| 270 | * @opcode: location to store the read opcode. | ||
| 271 | * | ||
| 272 | * Called with mm->mmap_sem held (for read and with a reference to | ||
| 273 | * mm. | ||
| 274 | * | ||
| 275 | * For mm @mm, read the opcode at @vaddr and store it in @opcode. | ||
| 276 | * Return 0 (success) or a negative errno. | ||
| 277 | */ | ||
| 278 | static int read_opcode(struct mm_struct *mm, unsigned long vaddr, uprobe_opcode_t *opcode) | ||
| 279 | { | ||
| 280 | struct page *page; | ||
| 281 | void *vaddr_new; | ||
| 282 | int ret; | ||
| 283 | |||
| 284 | ret = get_user_pages(NULL, mm, vaddr, 1, 0, 1, &page, NULL); | ||
| 285 | if (ret <= 0) | ||
| 286 | return ret; | ||
| 287 | |||
| 288 | vaddr_new = kmap_atomic(page); | ||
| 289 | vaddr &= ~PAGE_MASK; | ||
| 290 | memcpy(opcode, vaddr_new + vaddr, UPROBE_SWBP_INSN_SIZE); | ||
| 291 | kunmap_atomic(vaddr_new); | ||
| 292 | |||
| 293 | put_page(page); | ||
| 294 | |||
| 295 | return 0; | ||
| 296 | } | ||
| 297 | |||
| 298 | static int is_swbp_at_addr(struct mm_struct *mm, unsigned long vaddr) | ||
| 299 | { | ||
| 300 | uprobe_opcode_t opcode; | ||
| 301 | int result; | ||
| 302 | |||
| 303 | if (current->mm == mm) { | ||
| 304 | pagefault_disable(); | ||
| 305 | result = __copy_from_user_inatomic(&opcode, (void __user*)vaddr, | ||
| 306 | sizeof(opcode)); | ||
| 307 | pagefault_enable(); | ||
| 308 | |||
| 309 | if (likely(result == 0)) | ||
| 310 | goto out; | ||
| 311 | } | ||
| 312 | |||
| 313 | result = read_opcode(mm, vaddr, &opcode); | ||
| 314 | if (result) | ||
| 315 | return result; | ||
| 316 | out: | ||
| 317 | if (is_swbp_insn(&opcode)) | ||
| 318 | return 1; | ||
| 319 | |||
| 320 | return 0; | ||
| 321 | } | ||
| 322 | |||
| 323 | /** | ||
| 324 | * set_swbp - store breakpoint at a given address. | 299 | * set_swbp - store breakpoint at a given address. |
| 325 | * @auprobe: arch specific probepoint information. | 300 | * @auprobe: arch specific probepoint information. |
| 326 | * @mm: the probed process address space. | 301 | * @mm: the probed process address space. |
| @@ -331,18 +306,7 @@ out: | |||
| 331 | */ | 306 | */ |
| 332 | int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr) | 307 | int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr) |
| 333 | { | 308 | { |
| 334 | int result; | 309 | return write_opcode(mm, vaddr, UPROBE_SWBP_INSN); |
| 335 | /* | ||
| 336 | * See the comment near uprobes_hash(). | ||
| 337 | */ | ||
| 338 | result = is_swbp_at_addr(mm, vaddr); | ||
| 339 | if (result == 1) | ||
| 340 | return 0; | ||
| 341 | |||
| 342 | if (result) | ||
| 343 | return result; | ||
| 344 | |||
| 345 | return write_opcode(auprobe, mm, vaddr, UPROBE_SWBP_INSN); | ||
| 346 | } | 310 | } |
| 347 | 311 | ||
| 348 | /** | 312 | /** |
| @@ -357,16 +321,7 @@ int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned | |||
| 357 | int __weak | 321 | int __weak |
| 358 | set_orig_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr) | 322 | set_orig_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr) |
| 359 | { | 323 | { |
| 360 | int result; | 324 | return write_opcode(mm, vaddr, *(uprobe_opcode_t *)auprobe->insn); |
| 361 | |||
| 362 | result = is_swbp_at_addr(mm, vaddr); | ||
| 363 | if (!result) | ||
| 364 | return -EINVAL; | ||
| 365 | |||
| 366 | if (result != 1) | ||
| 367 | return result; | ||
| 368 | |||
| 369 | return write_opcode(auprobe, mm, vaddr, *(uprobe_opcode_t *)auprobe->insn); | ||
| 370 | } | 325 | } |
| 371 | 326 | ||
| 372 | static int match_uprobe(struct uprobe *l, struct uprobe *r) | 327 | static int match_uprobe(struct uprobe *l, struct uprobe *r) |
| @@ -473,7 +428,7 @@ static struct uprobe *insert_uprobe(struct uprobe *uprobe) | |||
| 473 | spin_unlock(&uprobes_treelock); | 428 | spin_unlock(&uprobes_treelock); |
| 474 | 429 | ||
| 475 | /* For now assume that the instruction need not be single-stepped */ | 430 | /* For now assume that the instruction need not be single-stepped */ |
| 476 | uprobe->flags |= UPROBE_SKIP_SSTEP; | 431 | __set_bit(UPROBE_SKIP_SSTEP, &uprobe->flags); |
| 477 | 432 | ||
| 478 | return u; | 433 | return u; |
| 479 | } | 434 | } |
| @@ -495,6 +450,7 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset) | |||
| 495 | uprobe->inode = igrab(inode); | 450 | uprobe->inode = igrab(inode); |
| 496 | uprobe->offset = offset; | 451 | uprobe->offset = offset; |
| 497 | init_rwsem(&uprobe->consumer_rwsem); | 452 | init_rwsem(&uprobe->consumer_rwsem); |
| 453 | mutex_init(&uprobe->copy_mutex); | ||
| 498 | 454 | ||
| 499 | /* add to uprobes_tree, sorted on inode:offset */ | 455 | /* add to uprobes_tree, sorted on inode:offset */ |
| 500 | cur_uprobe = insert_uprobe(uprobe); | 456 | cur_uprobe = insert_uprobe(uprobe); |
| @@ -515,7 +471,7 @@ static void handler_chain(struct uprobe *uprobe, struct pt_regs *regs) | |||
| 515 | { | 471 | { |
| 516 | struct uprobe_consumer *uc; | 472 | struct uprobe_consumer *uc; |
| 517 | 473 | ||
| 518 | if (!(uprobe->flags & UPROBE_RUN_HANDLER)) | 474 | if (!test_bit(UPROBE_RUN_HANDLER, &uprobe->flags)) |
| 519 | return; | 475 | return; |
| 520 | 476 | ||
| 521 | down_read(&uprobe->consumer_rwsem); | 477 | down_read(&uprobe->consumer_rwsem); |
| @@ -621,29 +577,43 @@ static int copy_insn(struct uprobe *uprobe, struct file *filp) | |||
| 621 | return __copy_insn(mapping, filp, uprobe->arch.insn, bytes, uprobe->offset); | 577 | return __copy_insn(mapping, filp, uprobe->arch.insn, bytes, uprobe->offset); |
| 622 | } | 578 | } |
| 623 | 579 | ||
| 624 | /* | 580 | static int prepare_uprobe(struct uprobe *uprobe, struct file *file, |
| 625 | * How mm->uprobes_state.count gets updated | 581 | struct mm_struct *mm, unsigned long vaddr) |
| 626 | * uprobe_mmap() increments the count if | 582 | { |
| 627 | * - it successfully adds a breakpoint. | 583 | int ret = 0; |
| 628 | * - it cannot add a breakpoint, but sees that there is a underlying | 584 | |
| 629 | * breakpoint (via a is_swbp_at_addr()). | 585 | if (test_bit(UPROBE_COPY_INSN, &uprobe->flags)) |
| 630 | * | 586 | return ret; |
| 631 | * uprobe_munmap() decrements the count if | 587 | |
| 632 | * - it sees a underlying breakpoint, (via is_swbp_at_addr) | 588 | mutex_lock(&uprobe->copy_mutex); |
| 633 | * (Subsequent uprobe_unregister wouldnt find the breakpoint | 589 | if (test_bit(UPROBE_COPY_INSN, &uprobe->flags)) |
| 634 | * unless a uprobe_mmap kicks in, since the old vma would be | 590 | goto out; |
| 635 | * dropped just after uprobe_munmap.) | 591 | |
| 636 | * | 592 | ret = copy_insn(uprobe, file); |
| 637 | * uprobe_register increments the count if: | 593 | if (ret) |
| 638 | * - it successfully adds a breakpoint. | 594 | goto out; |
| 639 | * | 595 | |
| 640 | * uprobe_unregister decrements the count if: | 596 | ret = -ENOTSUPP; |
| 641 | * - it sees a underlying breakpoint and removes successfully. | 597 | if (is_swbp_insn((uprobe_opcode_t *)uprobe->arch.insn)) |
| 642 | * (via is_swbp_at_addr) | 598 | goto out; |
| 643 | * (Subsequent uprobe_munmap wouldnt find the breakpoint | 599 | |
| 644 | * since there is no underlying breakpoint after the | 600 | ret = arch_uprobe_analyze_insn(&uprobe->arch, mm, vaddr); |
| 645 | * breakpoint removal.) | 601 | if (ret) |
| 646 | */ | 602 | goto out; |
| 603 | |||
| 604 | /* write_opcode() assumes we don't cross page boundary */ | ||
| 605 | BUG_ON((uprobe->offset & ~PAGE_MASK) + | ||
| 606 | UPROBE_SWBP_INSN_SIZE > PAGE_SIZE); | ||
| 607 | |||
| 608 | smp_wmb(); /* pairs with rmb() in find_active_uprobe() */ | ||
| 609 | set_bit(UPROBE_COPY_INSN, &uprobe->flags); | ||
| 610 | |||
| 611 | out: | ||
| 612 | mutex_unlock(&uprobe->copy_mutex); | ||
| 613 | |||
| 614 | return ret; | ||
| 615 | } | ||
| 616 | |||
| 647 | static int | 617 | static int |
| 648 | install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, | 618 | install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, |
| 649 | struct vm_area_struct *vma, unsigned long vaddr) | 619 | struct vm_area_struct *vma, unsigned long vaddr) |
| @@ -661,24 +631,9 @@ install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, | |||
| 661 | if (!uprobe->consumers) | 631 | if (!uprobe->consumers) |
| 662 | return 0; | 632 | return 0; |
| 663 | 633 | ||
| 664 | if (!(uprobe->flags & UPROBE_COPY_INSN)) { | 634 | ret = prepare_uprobe(uprobe, vma->vm_file, mm, vaddr); |
| 665 | ret = copy_insn(uprobe, vma->vm_file); | 635 | if (ret) |
| 666 | if (ret) | 636 | return ret; |
| 667 | return ret; | ||
| 668 | |||
| 669 | if (is_swbp_insn((uprobe_opcode_t *)uprobe->arch.insn)) | ||
| 670 | return -ENOTSUPP; | ||
| 671 | |||
| 672 | ret = arch_uprobe_analyze_insn(&uprobe->arch, mm, vaddr); | ||
| 673 | if (ret) | ||
| 674 | return ret; | ||
| 675 | |||
| 676 | /* write_opcode() assumes we don't cross page boundary */ | ||
| 677 | BUG_ON((uprobe->offset & ~PAGE_MASK) + | ||
| 678 | UPROBE_SWBP_INSN_SIZE > PAGE_SIZE); | ||
| 679 | |||
| 680 | uprobe->flags |= UPROBE_COPY_INSN; | ||
| 681 | } | ||
| 682 | 637 | ||
| 683 | /* | 638 | /* |
| 684 | * set MMF_HAS_UPROBES in advance for uprobe_pre_sstep_notifier(), | 639 | * set MMF_HAS_UPROBES in advance for uprobe_pre_sstep_notifier(), |
| @@ -697,15 +652,15 @@ install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, | |||
| 697 | return ret; | 652 | return ret; |
| 698 | } | 653 | } |
| 699 | 654 | ||
| 700 | static void | 655 | static int |
| 701 | remove_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, unsigned long vaddr) | 656 | remove_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, unsigned long vaddr) |
| 702 | { | 657 | { |
| 703 | /* can happen if uprobe_register() fails */ | 658 | /* can happen if uprobe_register() fails */ |
| 704 | if (!test_bit(MMF_HAS_UPROBES, &mm->flags)) | 659 | if (!test_bit(MMF_HAS_UPROBES, &mm->flags)) |
| 705 | return; | 660 | return 0; |
| 706 | 661 | ||
| 707 | set_bit(MMF_RECALC_UPROBES, &mm->flags); | 662 | set_bit(MMF_RECALC_UPROBES, &mm->flags); |
| 708 | set_orig_insn(&uprobe->arch, mm, vaddr); | 663 | return set_orig_insn(&uprobe->arch, mm, vaddr); |
| 709 | } | 664 | } |
| 710 | 665 | ||
| 711 | /* | 666 | /* |
| @@ -820,7 +775,7 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register) | |||
| 820 | struct mm_struct *mm = info->mm; | 775 | struct mm_struct *mm = info->mm; |
| 821 | struct vm_area_struct *vma; | 776 | struct vm_area_struct *vma; |
| 822 | 777 | ||
| 823 | if (err) | 778 | if (err && is_register) |
| 824 | goto free; | 779 | goto free; |
| 825 | 780 | ||
| 826 | down_write(&mm->mmap_sem); | 781 | down_write(&mm->mmap_sem); |
| @@ -836,7 +791,7 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register) | |||
| 836 | if (is_register) | 791 | if (is_register) |
| 837 | err = install_breakpoint(uprobe, mm, vma, info->vaddr); | 792 | err = install_breakpoint(uprobe, mm, vma, info->vaddr); |
| 838 | else | 793 | else |
| 839 | remove_breakpoint(uprobe, mm, info->vaddr); | 794 | err |= remove_breakpoint(uprobe, mm, info->vaddr); |
| 840 | 795 | ||
| 841 | unlock: | 796 | unlock: |
| 842 | up_write(&mm->mmap_sem); | 797 | up_write(&mm->mmap_sem); |
| @@ -893,13 +848,15 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer * | |||
| 893 | mutex_lock(uprobes_hash(inode)); | 848 | mutex_lock(uprobes_hash(inode)); |
| 894 | uprobe = alloc_uprobe(inode, offset); | 849 | uprobe = alloc_uprobe(inode, offset); |
| 895 | 850 | ||
| 896 | if (uprobe && !consumer_add(uprobe, uc)) { | 851 | if (!uprobe) { |
| 852 | ret = -ENOMEM; | ||
| 853 | } else if (!consumer_add(uprobe, uc)) { | ||
| 897 | ret = __uprobe_register(uprobe); | 854 | ret = __uprobe_register(uprobe); |
| 898 | if (ret) { | 855 | if (ret) { |
| 899 | uprobe->consumers = NULL; | 856 | uprobe->consumers = NULL; |
| 900 | __uprobe_unregister(uprobe); | 857 | __uprobe_unregister(uprobe); |
| 901 | } else { | 858 | } else { |
| 902 | uprobe->flags |= UPROBE_RUN_HANDLER; | 859 | set_bit(UPROBE_RUN_HANDLER, &uprobe->flags); |
| 903 | } | 860 | } |
| 904 | } | 861 | } |
| 905 | 862 | ||
| @@ -932,7 +889,7 @@ void uprobe_unregister(struct inode *inode, loff_t offset, struct uprobe_consume | |||
| 932 | if (consumer_del(uprobe, uc)) { | 889 | if (consumer_del(uprobe, uc)) { |
| 933 | if (!uprobe->consumers) { | 890 | if (!uprobe->consumers) { |
| 934 | __uprobe_unregister(uprobe); | 891 | __uprobe_unregister(uprobe); |
| 935 | uprobe->flags &= ~UPROBE_RUN_HANDLER; | 892 | clear_bit(UPROBE_RUN_HANDLER, &uprobe->flags); |
| 936 | } | 893 | } |
| 937 | } | 894 | } |
| 938 | 895 | ||
| @@ -1393,10 +1350,11 @@ bool uprobe_deny_signal(void) | |||
| 1393 | */ | 1350 | */ |
| 1394 | static bool can_skip_sstep(struct uprobe *uprobe, struct pt_regs *regs) | 1351 | static bool can_skip_sstep(struct uprobe *uprobe, struct pt_regs *regs) |
| 1395 | { | 1352 | { |
| 1396 | if (arch_uprobe_skip_sstep(&uprobe->arch, regs)) | 1353 | if (test_bit(UPROBE_SKIP_SSTEP, &uprobe->flags)) { |
| 1397 | return true; | 1354 | if (arch_uprobe_skip_sstep(&uprobe->arch, regs)) |
| 1398 | 1355 | return true; | |
| 1399 | uprobe->flags &= ~UPROBE_SKIP_SSTEP; | 1356 | clear_bit(UPROBE_SKIP_SSTEP, &uprobe->flags); |
| 1357 | } | ||
| 1400 | return false; | 1358 | return false; |
| 1401 | } | 1359 | } |
| 1402 | 1360 | ||
| @@ -1419,6 +1377,30 @@ static void mmf_recalc_uprobes(struct mm_struct *mm) | |||
| 1419 | clear_bit(MMF_HAS_UPROBES, &mm->flags); | 1377 | clear_bit(MMF_HAS_UPROBES, &mm->flags); |
| 1420 | } | 1378 | } |
| 1421 | 1379 | ||
| 1380 | static int is_swbp_at_addr(struct mm_struct *mm, unsigned long vaddr) | ||
| 1381 | { | ||
| 1382 | struct page *page; | ||
| 1383 | uprobe_opcode_t opcode; | ||
| 1384 | int result; | ||
| 1385 | |||
| 1386 | pagefault_disable(); | ||
| 1387 | result = __copy_from_user_inatomic(&opcode, (void __user*)vaddr, | ||
| 1388 | sizeof(opcode)); | ||
| 1389 | pagefault_enable(); | ||
| 1390 | |||
| 1391 | if (likely(result == 0)) | ||
| 1392 | goto out; | ||
| 1393 | |||
| 1394 | result = get_user_pages(NULL, mm, vaddr, 1, 0, 1, &page, NULL); | ||
| 1395 | if (result < 0) | ||
| 1396 | return result; | ||
| 1397 | |||
| 1398 | copy_opcode(page, vaddr, &opcode); | ||
| 1399 | put_page(page); | ||
| 1400 | out: | ||
| 1401 | return is_swbp_insn(&opcode); | ||
| 1402 | } | ||
| 1403 | |||
| 1422 | static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp) | 1404 | static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp) |
| 1423 | { | 1405 | { |
| 1424 | struct mm_struct *mm = current->mm; | 1406 | struct mm_struct *mm = current->mm; |
| @@ -1489,38 +1471,41 @@ static void handle_swbp(struct pt_regs *regs) | |||
| 1489 | } | 1471 | } |
| 1490 | return; | 1472 | return; |
| 1491 | } | 1473 | } |
| 1474 | /* | ||
| 1475 | * TODO: move copy_insn/etc into _register and remove this hack. | ||
| 1476 | * After we hit the bp, _unregister + _register can install the | ||
| 1477 | * new and not-yet-analyzed uprobe at the same address, restart. | ||
| 1478 | */ | ||
| 1479 | smp_rmb(); /* pairs with wmb() in install_breakpoint() */ | ||
| 1480 | if (unlikely(!test_bit(UPROBE_COPY_INSN, &uprobe->flags))) | ||
| 1481 | goto restart; | ||
| 1492 | 1482 | ||
| 1493 | utask = current->utask; | 1483 | utask = current->utask; |
| 1494 | if (!utask) { | 1484 | if (!utask) { |
| 1495 | utask = add_utask(); | 1485 | utask = add_utask(); |
| 1496 | /* Cannot allocate; re-execute the instruction. */ | 1486 | /* Cannot allocate; re-execute the instruction. */ |
| 1497 | if (!utask) | 1487 | if (!utask) |
| 1498 | goto cleanup_ret; | 1488 | goto restart; |
| 1499 | } | 1489 | } |
| 1500 | utask->active_uprobe = uprobe; | 1490 | |
| 1501 | handler_chain(uprobe, regs); | 1491 | handler_chain(uprobe, regs); |
| 1502 | if (uprobe->flags & UPROBE_SKIP_SSTEP && can_skip_sstep(uprobe, regs)) | 1492 | if (can_skip_sstep(uprobe, regs)) |
| 1503 | goto cleanup_ret; | 1493 | goto out; |
| 1504 | 1494 | ||
| 1505 | utask->state = UTASK_SSTEP; | ||
| 1506 | if (!pre_ssout(uprobe, regs, bp_vaddr)) { | 1495 | if (!pre_ssout(uprobe, regs, bp_vaddr)) { |
| 1507 | arch_uprobe_enable_step(&uprobe->arch); | 1496 | arch_uprobe_enable_step(&uprobe->arch); |
| 1497 | utask->active_uprobe = uprobe; | ||
| 1498 | utask->state = UTASK_SSTEP; | ||
| 1508 | return; | 1499 | return; |
| 1509 | } | 1500 | } |
| 1510 | 1501 | ||
| 1511 | cleanup_ret: | 1502 | restart: |
| 1512 | if (utask) { | 1503 | /* |
| 1513 | utask->active_uprobe = NULL; | 1504 | * cannot singlestep; cannot skip instruction; |
| 1514 | utask->state = UTASK_RUNNING; | 1505 | * re-execute the instruction. |
| 1515 | } | 1506 | */ |
| 1516 | if (!(uprobe->flags & UPROBE_SKIP_SSTEP)) | 1507 | instruction_pointer_set(regs, bp_vaddr); |
| 1517 | 1508 | out: | |
| 1518 | /* | ||
| 1519 | * cannot singlestep; cannot skip instruction; | ||
| 1520 | * re-execute the instruction. | ||
| 1521 | */ | ||
| 1522 | instruction_pointer_set(regs, bp_vaddr); | ||
| 1523 | |||
| 1524 | put_uprobe(uprobe); | 1509 | put_uprobe(uprobe); |
| 1525 | } | 1510 | } |
| 1526 | 1511 | ||
| @@ -1552,13 +1537,12 @@ static void handle_singlestep(struct uprobe_task *utask, struct pt_regs *regs) | |||
| 1552 | } | 1537 | } |
| 1553 | 1538 | ||
| 1554 | /* | 1539 | /* |
| 1555 | * On breakpoint hit, breakpoint notifier sets the TIF_UPROBE flag. (and on | 1540 | * On breakpoint hit, breakpoint notifier sets the TIF_UPROBE flag and |
| 1556 | * subsequent probe hits on the thread sets the state to UTASK_BP_HIT) and | 1541 | * allows the thread to return from interrupt. After that handle_swbp() |
| 1557 | * allows the thread to return from interrupt. | 1542 | * sets utask->active_uprobe. |
| 1558 | * | 1543 | * |
| 1559 | * On singlestep exception, singlestep notifier sets the TIF_UPROBE flag and | 1544 | * On singlestep exception, singlestep notifier sets the TIF_UPROBE flag |
| 1560 | * also sets the state to UTASK_SSTEP_ACK and allows the thread to return from | 1545 | * and allows the thread to return from interrupt. |
| 1561 | * interrupt. | ||
| 1562 | * | 1546 | * |
| 1563 | * While returning to userspace, thread notices the TIF_UPROBE flag and calls | 1547 | * While returning to userspace, thread notices the TIF_UPROBE flag and calls |
| 1564 | * uprobe_notify_resume(). | 1548 | * uprobe_notify_resume(). |
| @@ -1567,11 +1551,13 @@ void uprobe_notify_resume(struct pt_regs *regs) | |||
| 1567 | { | 1551 | { |
| 1568 | struct uprobe_task *utask; | 1552 | struct uprobe_task *utask; |
| 1569 | 1553 | ||
| 1554 | clear_thread_flag(TIF_UPROBE); | ||
| 1555 | |||
| 1570 | utask = current->utask; | 1556 | utask = current->utask; |
| 1571 | if (!utask || utask->state == UTASK_BP_HIT) | 1557 | if (utask && utask->active_uprobe) |
| 1572 | handle_swbp(regs); | ||
| 1573 | else | ||
| 1574 | handle_singlestep(utask, regs); | 1558 | handle_singlestep(utask, regs); |
| 1559 | else | ||
| 1560 | handle_swbp(regs); | ||
| 1575 | } | 1561 | } |
| 1576 | 1562 | ||
| 1577 | /* | 1563 | /* |
| @@ -1580,17 +1566,10 @@ void uprobe_notify_resume(struct pt_regs *regs) | |||
| 1580 | */ | 1566 | */ |
| 1581 | int uprobe_pre_sstep_notifier(struct pt_regs *regs) | 1567 | int uprobe_pre_sstep_notifier(struct pt_regs *regs) |
| 1582 | { | 1568 | { |
| 1583 | struct uprobe_task *utask; | ||
| 1584 | |||
| 1585 | if (!current->mm || !test_bit(MMF_HAS_UPROBES, ¤t->mm->flags)) | 1569 | if (!current->mm || !test_bit(MMF_HAS_UPROBES, ¤t->mm->flags)) |
| 1586 | return 0; | 1570 | return 0; |
| 1587 | 1571 | ||
| 1588 | utask = current->utask; | ||
| 1589 | if (utask) | ||
| 1590 | utask->state = UTASK_BP_HIT; | ||
| 1591 | |||
| 1592 | set_thread_flag(TIF_UPROBE); | 1572 | set_thread_flag(TIF_UPROBE); |
| 1593 | |||
| 1594 | return 1; | 1573 | return 1; |
| 1595 | } | 1574 | } |
| 1596 | 1575 | ||
diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index b32ed0e385a5..b979426d16c6 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c | |||
| @@ -1567,6 +1567,10 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size, | |||
| 1567 | 1567 | ||
| 1568 | put_online_cpus(); | 1568 | put_online_cpus(); |
| 1569 | } else { | 1569 | } else { |
| 1570 | /* Make sure this CPU has been intitialized */ | ||
| 1571 | if (!cpumask_test_cpu(cpu_id, buffer->cpumask)) | ||
| 1572 | goto out; | ||
| 1573 | |||
| 1570 | cpu_buffer = buffer->buffers[cpu_id]; | 1574 | cpu_buffer = buffer->buffers[cpu_id]; |
| 1571 | 1575 | ||
| 1572 | if (nr_pages == cpu_buffer->nr_pages) | 1576 | if (nr_pages == cpu_buffer->nr_pages) |
