diff options
Diffstat (limited to 'arch/um/os-Linux')
-rw-r--r-- | arch/um/os-Linux/Makefile | 14 | ||||
-rw-r--r-- | arch/um/os-Linux/aio.c | 414 | ||||
-rw-r--r-- | arch/um/os-Linux/elf_aux.c | 3 | ||||
-rw-r--r-- | arch/um/os-Linux/process.c | 58 | ||||
-rw-r--r-- | arch/um/os-Linux/start_up.c | 359 | ||||
-rw-r--r-- | arch/um/os-Linux/tt.c | 113 |
6 files changed, 954 insertions, 7 deletions
diff --git a/arch/um/os-Linux/Makefile b/arch/um/os-Linux/Makefile index 4ddf540284ce..7a1662419c0c 100644 --- a/arch/um/os-Linux/Makefile +++ b/arch/um/os-Linux/Makefile | |||
@@ -3,11 +3,19 @@ | |||
3 | # Licensed under the GPL | 3 | # Licensed under the GPL |
4 | # | 4 | # |
5 | 5 | ||
6 | obj-y = elf_aux.o file.o process.o signal.o time.o tty.o user_syms.o drivers/ \ | 6 | obj-y = aio.o elf_aux.o file.o process.o signal.o start_up.o time.o tt.o \ |
7 | sys-$(SUBARCH)/ | 7 | tty.o user_syms.o drivers/ sys-$(SUBARCH)/ |
8 | 8 | ||
9 | USER_OBJS := elf_aux.o file.o process.o signal.o time.o tty.o | 9 | USER_OBJS := aio.o elf_aux.o file.o process.o signal.o start_up.o time.o tt.o \ |
10 | tty.o | ||
11 | |||
12 | elf_aux.o: $(ARCH_DIR)/kernel-offsets.h | ||
13 | CFLAGS_elf_aux.o += -I$(objtree)/arch/um | ||
10 | 14 | ||
11 | CFLAGS_user_syms.o += -DSUBARCH_$(SUBARCH) | 15 | CFLAGS_user_syms.o += -DSUBARCH_$(SUBARCH) |
12 | 16 | ||
17 | HAVE_AIO_ABI := $(shell [ -r /usr/include/linux/aio_abi.h ] && \ | ||
18 | echo -DHAVE_AIO_ABI ) | ||
19 | CFLAGS_aio.o += $(HAVE_AIO_ABI) | ||
20 | |||
13 | include arch/um/scripts/Makefile.rules | 21 | include arch/um/scripts/Makefile.rules |
diff --git a/arch/um/os-Linux/aio.c b/arch/um/os-Linux/aio.c new file mode 100644 index 000000000000..b04897cd995d --- /dev/null +++ b/arch/um/os-Linux/aio.c | |||
@@ -0,0 +1,414 @@ | |||
1 | /* | ||
2 | * Copyright (C) 2004 Jeff Dike (jdike@addtoit.com) | ||
3 | * Licensed under the GPL | ||
4 | */ | ||
5 | |||
6 | #include <stdlib.h> | ||
7 | #include <unistd.h> | ||
8 | #include <signal.h> | ||
9 | #include <string.h> | ||
10 | #include <errno.h> | ||
11 | #include <sched.h> | ||
12 | #include <sys/syscall.h> | ||
13 | #include "os.h" | ||
14 | #include "helper.h" | ||
15 | #include "aio.h" | ||
16 | #include "init.h" | ||
17 | #include "user.h" | ||
18 | #include "mode.h" | ||
19 | |||
20 | static int aio_req_fd_r = -1; | ||
21 | static int aio_req_fd_w = -1; | ||
22 | |||
23 | static int update_aio(struct aio_context *aio, int res) | ||
24 | { | ||
25 | if(res < 0) | ||
26 | aio->len = res; | ||
27 | else if((res == 0) && (aio->type == AIO_READ)){ | ||
28 | /* This is the EOF case - we have hit the end of the file | ||
29 | * and it ends in a partial block, so we fill the end of | ||
30 | * the block with zeros and claim success. | ||
31 | */ | ||
32 | memset(aio->data, 0, aio->len); | ||
33 | aio->len = 0; | ||
34 | } | ||
35 | else if(res > 0){ | ||
36 | aio->len -= res; | ||
37 | aio->data += res; | ||
38 | aio->offset += res; | ||
39 | return aio->len; | ||
40 | } | ||
41 | |||
42 | return 0; | ||
43 | } | ||
44 | |||
45 | #if defined(HAVE_AIO_ABI) | ||
46 | #include <linux/aio_abi.h> | ||
47 | |||
48 | /* If we have the headers, we are going to build with AIO enabled. | ||
49 | * If we don't have aio in libc, we define the necessary stubs here. | ||
50 | */ | ||
51 | |||
52 | #if !defined(HAVE_AIO_LIBC) | ||
53 | |||
54 | static long io_setup(int n, aio_context_t *ctxp) | ||
55 | { | ||
56 | return syscall(__NR_io_setup, n, ctxp); | ||
57 | } | ||
58 | |||
59 | static long io_submit(aio_context_t ctx, long nr, struct iocb **iocbpp) | ||
60 | { | ||
61 | return syscall(__NR_io_submit, ctx, nr, iocbpp); | ||
62 | } | ||
63 | |||
64 | static long io_getevents(aio_context_t ctx_id, long min_nr, long nr, | ||
65 | struct io_event *events, struct timespec *timeout) | ||
66 | { | ||
67 | return syscall(__NR_io_getevents, ctx_id, min_nr, nr, events, timeout); | ||
68 | } | ||
69 | |||
70 | #endif | ||
71 | |||
72 | /* The AIO_MMAP cases force the mmapped page into memory here | ||
73 | * rather than in whatever place first touches the data. I used | ||
74 | * to do this by touching the page, but that's delicate because | ||
75 | * gcc is prone to optimizing that away. So, what's done here | ||
76 | * is we read from the descriptor from which the page was | ||
77 | * mapped. The caller is required to pass an offset which is | ||
78 | * inside the page that was mapped. Thus, when the read | ||
79 | * returns, we know that the page is in the page cache, and | ||
80 | * that it now backs the mmapped area. | ||
81 | */ | ||
82 | |||
83 | static int do_aio(aio_context_t ctx, struct aio_context *aio) | ||
84 | { | ||
85 | struct iocb iocb, *iocbp = &iocb; | ||
86 | char c; | ||
87 | int err; | ||
88 | |||
89 | iocb = ((struct iocb) { .aio_data = (unsigned long) aio, | ||
90 | .aio_reqprio = 0, | ||
91 | .aio_fildes = aio->fd, | ||
92 | .aio_buf = (unsigned long) aio->data, | ||
93 | .aio_nbytes = aio->len, | ||
94 | .aio_offset = aio->offset, | ||
95 | .aio_reserved1 = 0, | ||
96 | .aio_reserved2 = 0, | ||
97 | .aio_reserved3 = 0 }); | ||
98 | |||
99 | switch(aio->type){ | ||
100 | case AIO_READ: | ||
101 | iocb.aio_lio_opcode = IOCB_CMD_PREAD; | ||
102 | break; | ||
103 | case AIO_WRITE: | ||
104 | iocb.aio_lio_opcode = IOCB_CMD_PWRITE; | ||
105 | break; | ||
106 | case AIO_MMAP: | ||
107 | iocb.aio_lio_opcode = IOCB_CMD_PREAD; | ||
108 | iocb.aio_buf = (unsigned long) &c; | ||
109 | iocb.aio_nbytes = sizeof(c); | ||
110 | break; | ||
111 | default: | ||
112 | printk("Bogus op in do_aio - %d\n", aio->type); | ||
113 | err = -EINVAL; | ||
114 | goto out; | ||
115 | } | ||
116 | |||
117 | err = io_submit(ctx, 1, &iocbp); | ||
118 | if(err > 0) | ||
119 | err = 0; | ||
120 | |||
121 | out: | ||
122 | return err; | ||
123 | } | ||
124 | |||
125 | static aio_context_t ctx = 0; | ||
126 | |||
127 | static int aio_thread(void *arg) | ||
128 | { | ||
129 | struct aio_thread_reply reply; | ||
130 | struct aio_context *aio; | ||
131 | struct io_event event; | ||
132 | int err, n; | ||
133 | |||
134 | signal(SIGWINCH, SIG_IGN); | ||
135 | |||
136 | while(1){ | ||
137 | n = io_getevents(ctx, 1, 1, &event, NULL); | ||
138 | if(n < 0){ | ||
139 | if(errno == EINTR) | ||
140 | continue; | ||
141 | printk("aio_thread - io_getevents failed, " | ||
142 | "errno = %d\n", errno); | ||
143 | } | ||
144 | else { | ||
145 | aio = (struct aio_context *) event.data; | ||
146 | if(update_aio(aio, event.res)){ | ||
147 | do_aio(ctx, aio); | ||
148 | continue; | ||
149 | } | ||
150 | |||
151 | reply = ((struct aio_thread_reply) | ||
152 | { .data = aio, | ||
153 | .err = aio->len }); | ||
154 | err = os_write_file(aio->reply_fd, &reply, | ||
155 | sizeof(reply)); | ||
156 | if(err != sizeof(reply)) | ||
157 | printk("aio_thread - write failed, " | ||
158 | "fd = %d, err = %d\n", aio->reply_fd, | ||
159 | -err); | ||
160 | } | ||
161 | } | ||
162 | return 0; | ||
163 | } | ||
164 | |||
165 | #endif | ||
166 | |||
167 | static int do_not_aio(struct aio_context *aio) | ||
168 | { | ||
169 | char c; | ||
170 | int err; | ||
171 | |||
172 | switch(aio->type){ | ||
173 | case AIO_READ: | ||
174 | err = os_seek_file(aio->fd, aio->offset); | ||
175 | if(err) | ||
176 | goto out; | ||
177 | |||
178 | err = os_read_file(aio->fd, aio->data, aio->len); | ||
179 | break; | ||
180 | case AIO_WRITE: | ||
181 | err = os_seek_file(aio->fd, aio->offset); | ||
182 | if(err) | ||
183 | goto out; | ||
184 | |||
185 | err = os_write_file(aio->fd, aio->data, aio->len); | ||
186 | break; | ||
187 | case AIO_MMAP: | ||
188 | err = os_seek_file(aio->fd, aio->offset); | ||
189 | if(err) | ||
190 | goto out; | ||
191 | |||
192 | err = os_read_file(aio->fd, &c, sizeof(c)); | ||
193 | break; | ||
194 | default: | ||
195 | printk("do_not_aio - bad request type : %d\n", aio->type); | ||
196 | err = -EINVAL; | ||
197 | break; | ||
198 | } | ||
199 | |||
200 | out: | ||
201 | return err; | ||
202 | } | ||
203 | |||
204 | static int not_aio_thread(void *arg) | ||
205 | { | ||
206 | struct aio_context *aio; | ||
207 | struct aio_thread_reply reply; | ||
208 | int err; | ||
209 | |||
210 | signal(SIGWINCH, SIG_IGN); | ||
211 | while(1){ | ||
212 | err = os_read_file(aio_req_fd_r, &aio, sizeof(aio)); | ||
213 | if(err != sizeof(aio)){ | ||
214 | if(err < 0) | ||
215 | printk("not_aio_thread - read failed, " | ||
216 | "fd = %d, err = %d\n", aio_req_fd_r, | ||
217 | -err); | ||
218 | else { | ||
219 | printk("not_aio_thread - short read, fd = %d, " | ||
220 | "length = %d\n", aio_req_fd_r, err); | ||
221 | } | ||
222 | continue; | ||
223 | } | ||
224 | again: | ||
225 | err = do_not_aio(aio); | ||
226 | |||
227 | if(update_aio(aio, err)) | ||
228 | goto again; | ||
229 | |||
230 | reply = ((struct aio_thread_reply) { .data = aio, | ||
231 | .err = aio->len }); | ||
232 | err = os_write_file(aio->reply_fd, &reply, sizeof(reply)); | ||
233 | if(err != sizeof(reply)) | ||
234 | printk("not_aio_thread - write failed, fd = %d, " | ||
235 | "err = %d\n", aio_req_fd_r, -err); | ||
236 | } | ||
237 | } | ||
238 | |||
239 | static int submit_aio_24(struct aio_context *aio) | ||
240 | { | ||
241 | int err; | ||
242 | |||
243 | err = os_write_file(aio_req_fd_w, &aio, sizeof(aio)); | ||
244 | if(err == sizeof(aio)) | ||
245 | err = 0; | ||
246 | |||
247 | return err; | ||
248 | } | ||
249 | |||
250 | static int aio_pid = -1; | ||
251 | static int (*submit_proc)(struct aio_context *aio); | ||
252 | |||
253 | static int init_aio_24(void) | ||
254 | { | ||
255 | unsigned long stack; | ||
256 | int fds[2], err; | ||
257 | |||
258 | err = os_pipe(fds, 1, 1); | ||
259 | if(err) | ||
260 | goto out; | ||
261 | |||
262 | aio_req_fd_w = fds[0]; | ||
263 | aio_req_fd_r = fds[1]; | ||
264 | err = run_helper_thread(not_aio_thread, NULL, | ||
265 | CLONE_FILES | CLONE_VM | SIGCHLD, &stack, 0); | ||
266 | if(err < 0) | ||
267 | goto out_close_pipe; | ||
268 | |||
269 | aio_pid = err; | ||
270 | goto out; | ||
271 | |||
272 | out_close_pipe: | ||
273 | os_close_file(fds[0]); | ||
274 | os_close_file(fds[1]); | ||
275 | aio_req_fd_w = -1; | ||
276 | aio_req_fd_r = -1; | ||
277 | out: | ||
278 | #ifndef HAVE_AIO_ABI | ||
279 | printk("/usr/include/linux/aio_abi.h not present during build\n"); | ||
280 | #endif | ||
281 | printk("2.6 host AIO support not used - falling back to I/O " | ||
282 | "thread\n"); | ||
283 | |||
284 | submit_proc = submit_aio_24; | ||
285 | |||
286 | return 0; | ||
287 | } | ||
288 | |||
289 | #ifdef HAVE_AIO_ABI | ||
290 | #define DEFAULT_24_AIO 0 | ||
291 | static int submit_aio_26(struct aio_context *aio) | ||
292 | { | ||
293 | struct aio_thread_reply reply; | ||
294 | int err; | ||
295 | |||
296 | err = do_aio(ctx, aio); | ||
297 | if(err){ | ||
298 | reply = ((struct aio_thread_reply) { .data = aio, | ||
299 | .err = err }); | ||
300 | err = os_write_file(aio->reply_fd, &reply, sizeof(reply)); | ||
301 | if(err != sizeof(reply)) | ||
302 | printk("submit_aio_26 - write failed, " | ||
303 | "fd = %d, err = %d\n", aio->reply_fd, -err); | ||
304 | else err = 0; | ||
305 | } | ||
306 | |||
307 | return err; | ||
308 | } | ||
309 | |||
310 | static int init_aio_26(void) | ||
311 | { | ||
312 | unsigned long stack; | ||
313 | int err; | ||
314 | |||
315 | if(io_setup(256, &ctx)){ | ||
316 | printk("aio_thread failed to initialize context, err = %d\n", | ||
317 | errno); | ||
318 | return -errno; | ||
319 | } | ||
320 | |||
321 | err = run_helper_thread(aio_thread, NULL, | ||
322 | CLONE_FILES | CLONE_VM | SIGCHLD, &stack, 0); | ||
323 | if(err < 0) | ||
324 | return -errno; | ||
325 | |||
326 | aio_pid = err; | ||
327 | |||
328 | printk("Using 2.6 host AIO\n"); | ||
329 | |||
330 | submit_proc = submit_aio_26; | ||
331 | |||
332 | return 0; | ||
333 | } | ||
334 | |||
335 | #else | ||
336 | #define DEFAULT_24_AIO 1 | ||
337 | static int submit_aio_26(struct aio_context *aio) | ||
338 | { | ||
339 | return -ENOSYS; | ||
340 | } | ||
341 | |||
342 | static int init_aio_26(void) | ||
343 | { | ||
344 | submit_proc = submit_aio_26; | ||
345 | return -ENOSYS; | ||
346 | } | ||
347 | #endif | ||
348 | |||
349 | static int aio_24 = DEFAULT_24_AIO; | ||
350 | |||
351 | static int __init set_aio_24(char *name, int *add) | ||
352 | { | ||
353 | aio_24 = 1; | ||
354 | return 0; | ||
355 | } | ||
356 | |||
357 | __uml_setup("aio=2.4", set_aio_24, | ||
358 | "aio=2.4\n" | ||
359 | " This is used to force UML to use 2.4-style AIO even when 2.6 AIO is\n" | ||
360 | " available. 2.4 AIO is a single thread that handles one request at a\n" | ||
361 | " time, synchronously. 2.6 AIO is a thread which uses the 2.6 AIO \n" | ||
362 | " interface to handle an arbitrary number of pending requests. 2.6 AIO \n" | ||
363 | " is not available in tt mode, on 2.4 hosts, or when UML is built with\n" | ||
364 | " /usr/include/linux/aio_abi.h not available. Many distributions don't\n" | ||
365 | " include aio_abi.h, so you will need to copy it from a kernel tree to\n" | ||
366 | " your /usr/include/linux in order to build an AIO-capable UML\n\n" | ||
367 | ); | ||
368 | |||
369 | static int init_aio(void) | ||
370 | { | ||
371 | int err; | ||
372 | |||
373 | CHOOSE_MODE(({ | ||
374 | if(!aio_24){ | ||
375 | printk("Disabling 2.6 AIO in tt mode\n"); | ||
376 | aio_24 = 1; | ||
377 | } }), (void) 0); | ||
378 | |||
379 | if(!aio_24){ | ||
380 | err = init_aio_26(); | ||
381 | if(err && (errno == ENOSYS)){ | ||
382 | printk("2.6 AIO not supported on the host - " | ||
383 | "reverting to 2.4 AIO\n"); | ||
384 | aio_24 = 1; | ||
385 | } | ||
386 | else return err; | ||
387 | } | ||
388 | |||
389 | if(aio_24) | ||
390 | return init_aio_24(); | ||
391 | |||
392 | return 0; | ||
393 | } | ||
394 | |||
395 | /* The reason for the __initcall/__uml_exitcall asymmetry is that init_aio | ||
396 | * needs to be called when the kernel is running because it calls run_helper, | ||
397 | * which needs get_free_page. exit_aio is a __uml_exitcall because the generic | ||
398 | * kernel does not run __exitcalls on shutdown, and can't because many of them | ||
399 | * break when called outside of module unloading. | ||
400 | */ | ||
401 | __initcall(init_aio); | ||
402 | |||
403 | static void exit_aio(void) | ||
404 | { | ||
405 | if(aio_pid != -1) | ||
406 | os_kill_process(aio_pid, 1); | ||
407 | } | ||
408 | |||
409 | __uml_exitcall(exit_aio); | ||
410 | |||
411 | int submit_aio(struct aio_context *aio) | ||
412 | { | ||
413 | return (*submit_proc)(aio); | ||
414 | } | ||
diff --git a/arch/um/os-Linux/elf_aux.c b/arch/um/os-Linux/elf_aux.c index 4cca3e9c23fe..1399520a8588 100644 --- a/arch/um/os-Linux/elf_aux.c +++ b/arch/um/os-Linux/elf_aux.c | |||
@@ -12,8 +12,9 @@ | |||
12 | #include "init.h" | 12 | #include "init.h" |
13 | #include "elf_user.h" | 13 | #include "elf_user.h" |
14 | #include "mem_user.h" | 14 | #include "mem_user.h" |
15 | #include <kernel-offsets.h> | ||
15 | 16 | ||
16 | #if ELF_CLASS == ELFCLASS32 | 17 | #if HOST_ELF_CLASS == ELFCLASS32 |
17 | typedef Elf32_auxv_t elf_auxv_t; | 18 | typedef Elf32_auxv_t elf_auxv_t; |
18 | #else | 19 | #else |
19 | typedef Elf64_auxv_t elf_auxv_t; | 20 | typedef Elf64_auxv_t elf_auxv_t; |
diff --git a/arch/um/os-Linux/process.c b/arch/um/os-Linux/process.c index 1e126bfd31a7..d32413e4b4ce 100644 --- a/arch/um/os-Linux/process.c +++ b/arch/um/os-Linux/process.c | |||
@@ -3,10 +3,10 @@ | |||
3 | * Licensed under the GPL | 3 | * Licensed under the GPL |
4 | */ | 4 | */ |
5 | 5 | ||
6 | #include <unistd.h> | ||
7 | #include <stdio.h> | 6 | #include <stdio.h> |
8 | #include <errno.h> | 7 | #include <errno.h> |
9 | #include <signal.h> | 8 | #include <signal.h> |
9 | #include <setjmp.h> | ||
10 | #include <linux/unistd.h> | 10 | #include <linux/unistd.h> |
11 | #include <sys/mman.h> | 11 | #include <sys/mman.h> |
12 | #include <sys/wait.h> | 12 | #include <sys/wait.h> |
@@ -14,6 +14,10 @@ | |||
14 | #include "os.h" | 14 | #include "os.h" |
15 | #include "user.h" | 15 | #include "user.h" |
16 | #include "user_util.h" | 16 | #include "user_util.h" |
17 | #include "signal_user.h" | ||
18 | #include "process.h" | ||
19 | #include "irq_user.h" | ||
20 | #include "kern_util.h" | ||
17 | 21 | ||
18 | #define ARBITRARY_ADDR -1 | 22 | #define ARBITRARY_ADDR -1 |
19 | #define FAILURE_PID -1 | 23 | #define FAILURE_PID -1 |
@@ -114,8 +118,10 @@ void os_usr1_process(int pid) | |||
114 | kill(pid, SIGUSR1); | 118 | kill(pid, SIGUSR1); |
115 | } | 119 | } |
116 | 120 | ||
117 | /*Don't use the glibc version, which caches the result in TLS. It misses some | 121 | /* Don't use the glibc version, which caches the result in TLS. It misses some |
118 | * syscalls, and also breaks with clone(), which does not unshare the TLS.*/ | 122 | * syscalls, and also breaks with clone(), which does not unshare the TLS. |
123 | */ | ||
124 | |||
119 | inline _syscall0(pid_t, getpid) | 125 | inline _syscall0(pid_t, getpid) |
120 | 126 | ||
121 | int os_getpid(void) | 127 | int os_getpid(void) |
@@ -164,6 +170,52 @@ int os_unmap_memory(void *addr, int len) | |||
164 | return(0); | 170 | return(0); |
165 | } | 171 | } |
166 | 172 | ||
173 | void init_new_thread_stack(void *sig_stack, void (*usr1_handler)(int)) | ||
174 | { | ||
175 | int flags = 0, pages; | ||
176 | |||
177 | if(sig_stack != NULL){ | ||
178 | pages = (1 << UML_CONFIG_KERNEL_STACK_ORDER); | ||
179 | set_sigstack(sig_stack, pages * page_size()); | ||
180 | flags = SA_ONSTACK; | ||
181 | } | ||
182 | if(usr1_handler) set_handler(SIGUSR1, usr1_handler, flags, -1); | ||
183 | } | ||
184 | |||
185 | void init_new_thread_signals(int altstack) | ||
186 | { | ||
187 | int flags = altstack ? SA_ONSTACK : 0; | ||
188 | |||
189 | set_handler(SIGSEGV, (__sighandler_t) sig_handler, flags, | ||
190 | SIGUSR1, SIGIO, SIGWINCH, SIGALRM, SIGVTALRM, -1); | ||
191 | set_handler(SIGTRAP, (__sighandler_t) sig_handler, flags, | ||
192 | SIGUSR1, SIGIO, SIGWINCH, SIGALRM, SIGVTALRM, -1); | ||
193 | set_handler(SIGFPE, (__sighandler_t) sig_handler, flags, | ||
194 | SIGUSR1, SIGIO, SIGWINCH, SIGALRM, SIGVTALRM, -1); | ||
195 | set_handler(SIGILL, (__sighandler_t) sig_handler, flags, | ||
196 | SIGUSR1, SIGIO, SIGWINCH, SIGALRM, SIGVTALRM, -1); | ||
197 | set_handler(SIGBUS, (__sighandler_t) sig_handler, flags, | ||
198 | SIGUSR1, SIGIO, SIGWINCH, SIGALRM, SIGVTALRM, -1); | ||
199 | set_handler(SIGUSR2, (__sighandler_t) sig_handler, | ||
200 | flags, SIGUSR1, SIGIO, SIGWINCH, SIGALRM, SIGVTALRM, -1); | ||
201 | signal(SIGHUP, SIG_IGN); | ||
202 | |||
203 | init_irq_signals(altstack); | ||
204 | } | ||
205 | |||
206 | int run_kernel_thread(int (*fn)(void *), void *arg, void **jmp_ptr) | ||
207 | { | ||
208 | sigjmp_buf buf; | ||
209 | int n; | ||
210 | |||
211 | *jmp_ptr = &buf; | ||
212 | n = sigsetjmp(buf, 1); | ||
213 | if(n != 0) | ||
214 | return(n); | ||
215 | (*fn)(arg); | ||
216 | return(0); | ||
217 | } | ||
218 | |||
167 | /* | 219 | /* |
168 | * Overrides for Emacs so that we follow Linus's tabbing style. | 220 | * Overrides for Emacs so that we follow Linus's tabbing style. |
169 | * Emacs will notice this stuff at the end of the file and automatically | 221 | * Emacs will notice this stuff at the end of the file and automatically |
diff --git a/arch/um/os-Linux/start_up.c b/arch/um/os-Linux/start_up.c new file mode 100644 index 000000000000..040cc1472bc7 --- /dev/null +++ b/arch/um/os-Linux/start_up.c | |||
@@ -0,0 +1,359 @@ | |||
1 | /* | ||
2 | * Copyright (C) 2000, 2001, 2002 Jeff Dike (jdike@karaya.com) | ||
3 | * Licensed under the GPL | ||
4 | */ | ||
5 | |||
6 | #include <stdio.h> | ||
7 | #include <unistd.h> | ||
8 | #include <signal.h> | ||
9 | #include <sched.h> | ||
10 | #include <errno.h> | ||
11 | #include <stdarg.h> | ||
12 | #include <stdlib.h> | ||
13 | #include <setjmp.h> | ||
14 | #include <sys/time.h> | ||
15 | #include <sys/wait.h> | ||
16 | #include <sys/mman.h> | ||
17 | #include <asm/unistd.h> | ||
18 | #include <asm/page.h> | ||
19 | #include "user_util.h" | ||
20 | #include "kern_util.h" | ||
21 | #include "user.h" | ||
22 | #include "signal_kern.h" | ||
23 | #include "signal_user.h" | ||
24 | #include "sysdep/ptrace.h" | ||
25 | #include "sysdep/sigcontext.h" | ||
26 | #include "irq_user.h" | ||
27 | #include "ptrace_user.h" | ||
28 | #include "time_user.h" | ||
29 | #include "init.h" | ||
30 | #include "os.h" | ||
31 | #include "uml-config.h" | ||
32 | #include "choose-mode.h" | ||
33 | #include "mode.h" | ||
34 | #include "tempfile.h" | ||
35 | #ifdef UML_CONFIG_MODE_SKAS | ||
36 | #include "skas.h" | ||
37 | #include "skas_ptrace.h" | ||
38 | #include "registers.h" | ||
39 | #endif | ||
40 | |||
41 | static int ptrace_child(void *arg) | ||
42 | { | ||
43 | int ret; | ||
44 | int pid = os_getpid(), ppid = getppid(); | ||
45 | int sc_result; | ||
46 | |||
47 | if(ptrace(PTRACE_TRACEME, 0, 0, 0) < 0){ | ||
48 | perror("ptrace"); | ||
49 | os_kill_process(pid, 0); | ||
50 | } | ||
51 | os_stop_process(pid); | ||
52 | |||
53 | /*This syscall will be intercepted by the parent. Don't call more than | ||
54 | * once, please.*/ | ||
55 | sc_result = os_getpid(); | ||
56 | |||
57 | if (sc_result == pid) | ||
58 | ret = 1; /*Nothing modified by the parent, we are running | ||
59 | normally.*/ | ||
60 | else if (sc_result == ppid) | ||
61 | ret = 0; /*Expected in check_ptrace and check_sysemu when they | ||
62 | succeed in modifying the stack frame*/ | ||
63 | else | ||
64 | ret = 2; /*Serious trouble! This could be caused by a bug in | ||
65 | host 2.6 SKAS3/2.6 patch before release -V6, together | ||
66 | with a bug in the UML code itself.*/ | ||
67 | _exit(ret); | ||
68 | } | ||
69 | |||
70 | static int start_ptraced_child(void **stack_out) | ||
71 | { | ||
72 | void *stack; | ||
73 | unsigned long sp; | ||
74 | int pid, n, status; | ||
75 | |||
76 | stack = mmap(NULL, PAGE_SIZE, PROT_READ | PROT_WRITE | PROT_EXEC, | ||
77 | MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); | ||
78 | if(stack == MAP_FAILED) | ||
79 | panic("check_ptrace : mmap failed, errno = %d", errno); | ||
80 | sp = (unsigned long) stack + PAGE_SIZE - sizeof(void *); | ||
81 | pid = clone(ptrace_child, (void *) sp, SIGCHLD, NULL); | ||
82 | if(pid < 0) | ||
83 | panic("start_ptraced_child : clone failed, errno = %d", errno); | ||
84 | CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED)); | ||
85 | if(n < 0) | ||
86 | panic("check_ptrace : clone failed, errno = %d", errno); | ||
87 | if(!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGSTOP)) | ||
88 | panic("check_ptrace : expected SIGSTOP, got status = %d", | ||
89 | status); | ||
90 | |||
91 | *stack_out = stack; | ||
92 | return(pid); | ||
93 | } | ||
94 | |||
95 | /* When testing for SYSEMU support, if it is one of the broken versions, we | ||
96 | * must just avoid using sysemu, not panic, but only if SYSEMU features are | ||
97 | * broken. | ||
98 | * So only for SYSEMU features we test mustpanic, while normal host features | ||
99 | * must work anyway! | ||
100 | */ | ||
101 | static int stop_ptraced_child(int pid, void *stack, int exitcode, | ||
102 | int mustpanic) | ||
103 | { | ||
104 | int status, n, ret = 0; | ||
105 | |||
106 | if(ptrace(PTRACE_CONT, pid, 0, 0) < 0) | ||
107 | panic("check_ptrace : ptrace failed, errno = %d", errno); | ||
108 | CATCH_EINTR(n = waitpid(pid, &status, 0)); | ||
109 | if(!WIFEXITED(status) || (WEXITSTATUS(status) != exitcode)) { | ||
110 | int exit_with = WEXITSTATUS(status); | ||
111 | if (exit_with == 2) | ||
112 | printk("check_ptrace : child exited with status 2. " | ||
113 | "Serious trouble happening! Try updating your " | ||
114 | "host skas patch!\nDisabling SYSEMU support."); | ||
115 | printk("check_ptrace : child exited with exitcode %d, while " | ||
116 | "expecting %d; status 0x%x", exit_with, | ||
117 | exitcode, status); | ||
118 | if (mustpanic) | ||
119 | panic("\n"); | ||
120 | else | ||
121 | printk("\n"); | ||
122 | ret = -1; | ||
123 | } | ||
124 | |||
125 | if(munmap(stack, PAGE_SIZE) < 0) | ||
126 | panic("check_ptrace : munmap failed, errno = %d", errno); | ||
127 | return ret; | ||
128 | } | ||
129 | |||
130 | int ptrace_faultinfo = 1; | ||
131 | int proc_mm = 1; | ||
132 | |||
133 | static int __init skas0_cmd_param(char *str, int* add) | ||
134 | { | ||
135 | ptrace_faultinfo = proc_mm = 0; | ||
136 | return 0; | ||
137 | } | ||
138 | |||
139 | __uml_setup("skas0", skas0_cmd_param, | ||
140 | "skas0\n" | ||
141 | " Disables SKAS3 usage, so that SKAS0 is used, unless \n" | ||
142 | " you specify mode=tt.\n\n"); | ||
143 | |||
144 | static int force_sysemu_disabled = 0; | ||
145 | |||
146 | static int __init nosysemu_cmd_param(char *str, int* add) | ||
147 | { | ||
148 | force_sysemu_disabled = 1; | ||
149 | return 0; | ||
150 | } | ||
151 | |||
152 | __uml_setup("nosysemu", nosysemu_cmd_param, | ||
153 | "nosysemu\n" | ||
154 | " Turns off syscall emulation patch for ptrace (SYSEMU) on.\n" | ||
155 | " SYSEMU is a performance-patch introduced by Laurent Vivier. It changes\n" | ||
156 | " behaviour of ptrace() and helps reducing host context switch rate.\n" | ||
157 | " To make it working, you need a kernel patch for your host, too.\n" | ||
158 | " See http://perso.wanadoo.fr/laurent.vivier/UML/ for further \n" | ||
159 | " information.\n\n"); | ||
160 | |||
161 | static void __init check_sysemu(void) | ||
162 | { | ||
163 | void *stack; | ||
164 | int pid, n, status, count=0; | ||
165 | |||
166 | printk("Checking syscall emulation patch for ptrace..."); | ||
167 | sysemu_supported = 0; | ||
168 | pid = start_ptraced_child(&stack); | ||
169 | |||
170 | if(ptrace(PTRACE_SYSEMU, pid, 0, 0) < 0) | ||
171 | goto fail; | ||
172 | |||
173 | CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED)); | ||
174 | if (n < 0) | ||
175 | panic("check_sysemu : wait failed, errno = %d", errno); | ||
176 | if(!WIFSTOPPED(status) || (WSTOPSIG(status) != SIGTRAP)) | ||
177 | panic("check_sysemu : expected SIGTRAP, " | ||
178 | "got status = %d", status); | ||
179 | |||
180 | n = ptrace(PTRACE_POKEUSR, pid, PT_SYSCALL_RET_OFFSET, | ||
181 | os_getpid()); | ||
182 | if(n < 0) | ||
183 | panic("check_sysemu : failed to modify system " | ||
184 | "call return, errno = %d", errno); | ||
185 | |||
186 | if (stop_ptraced_child(pid, stack, 0, 0) < 0) | ||
187 | goto fail_stopped; | ||
188 | |||
189 | sysemu_supported = 1; | ||
190 | printk("OK\n"); | ||
191 | set_using_sysemu(!force_sysemu_disabled); | ||
192 | |||
193 | printk("Checking advanced syscall emulation patch for ptrace..."); | ||
194 | pid = start_ptraced_child(&stack); | ||
195 | |||
196 | if(ptrace(PTRACE_OLDSETOPTIONS, pid, 0, | ||
197 | (void *) PTRACE_O_TRACESYSGOOD) < 0) | ||
198 | panic("check_ptrace: PTRACE_OLDSETOPTIONS failed, errno = %d", | ||
199 | errno); | ||
200 | |||
201 | while(1){ | ||
202 | count++; | ||
203 | if(ptrace(PTRACE_SYSEMU_SINGLESTEP, pid, 0, 0) < 0) | ||
204 | goto fail; | ||
205 | CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED)); | ||
206 | if(n < 0) | ||
207 | panic("check_ptrace : wait failed, errno = %d", errno); | ||
208 | if(WIFSTOPPED(status) && (WSTOPSIG(status) == (SIGTRAP|0x80))){ | ||
209 | if (!count) | ||
210 | panic("check_ptrace : SYSEMU_SINGLESTEP " | ||
211 | "doesn't singlestep"); | ||
212 | n = ptrace(PTRACE_POKEUSR, pid, PT_SYSCALL_RET_OFFSET, | ||
213 | os_getpid()); | ||
214 | if(n < 0) | ||
215 | panic("check_sysemu : failed to modify system " | ||
216 | "call return, errno = %d", errno); | ||
217 | break; | ||
218 | } | ||
219 | else if(WIFSTOPPED(status) && (WSTOPSIG(status) == SIGTRAP)) | ||
220 | count++; | ||
221 | else | ||
222 | panic("check_ptrace : expected SIGTRAP or " | ||
223 | "(SIGTRAP|0x80), got status = %d", status); | ||
224 | } | ||
225 | if (stop_ptraced_child(pid, stack, 0, 0) < 0) | ||
226 | goto fail_stopped; | ||
227 | |||
228 | sysemu_supported = 2; | ||
229 | printk("OK\n"); | ||
230 | |||
231 | if ( !force_sysemu_disabled ) | ||
232 | set_using_sysemu(sysemu_supported); | ||
233 | return; | ||
234 | |||
235 | fail: | ||
236 | stop_ptraced_child(pid, stack, 1, 0); | ||
237 | fail_stopped: | ||
238 | printk("missing\n"); | ||
239 | } | ||
240 | |||
241 | static void __init check_ptrace(void) | ||
242 | { | ||
243 | void *stack; | ||
244 | int pid, syscall, n, status; | ||
245 | |||
246 | printk("Checking that ptrace can change system call numbers..."); | ||
247 | pid = start_ptraced_child(&stack); | ||
248 | |||
249 | if(ptrace(PTRACE_OLDSETOPTIONS, pid, 0, (void *)PTRACE_O_TRACESYSGOOD) < 0) | ||
250 | panic("check_ptrace: PTRACE_OLDSETOPTIONS failed, errno = %d", errno); | ||
251 | |||
252 | while(1){ | ||
253 | if(ptrace(PTRACE_SYSCALL, pid, 0, 0) < 0) | ||
254 | panic("check_ptrace : ptrace failed, errno = %d", | ||
255 | errno); | ||
256 | CATCH_EINTR(n = waitpid(pid, &status, WUNTRACED)); | ||
257 | if(n < 0) | ||
258 | panic("check_ptrace : wait failed, errno = %d", errno); | ||
259 | if(!WIFSTOPPED(status) || (WSTOPSIG(status) != (SIGTRAP|0x80))) | ||
260 | panic("check_ptrace : expected (SIGTRAP|0x80), " | ||
261 | "got status = %d", status); | ||
262 | |||
263 | syscall = ptrace(PTRACE_PEEKUSR, pid, PT_SYSCALL_NR_OFFSET, | ||
264 | 0); | ||
265 | if(syscall == __NR_getpid){ | ||
266 | n = ptrace(PTRACE_POKEUSR, pid, PT_SYSCALL_NR_OFFSET, | ||
267 | __NR_getppid); | ||
268 | if(n < 0) | ||
269 | panic("check_ptrace : failed to modify system " | ||
270 | "call, errno = %d", errno); | ||
271 | break; | ||
272 | } | ||
273 | } | ||
274 | stop_ptraced_child(pid, stack, 0, 1); | ||
275 | printk("OK\n"); | ||
276 | check_sysemu(); | ||
277 | } | ||
278 | |||
279 | void os_early_checks(void) | ||
280 | { | ||
281 | check_ptrace(); | ||
282 | } | ||
283 | |||
284 | static int __init noprocmm_cmd_param(char *str, int* add) | ||
285 | { | ||
286 | proc_mm = 0; | ||
287 | return 0; | ||
288 | } | ||
289 | |||
290 | __uml_setup("noprocmm", noprocmm_cmd_param, | ||
291 | "noprocmm\n" | ||
292 | " Turns off usage of /proc/mm, even if host supports it.\n" | ||
293 | " To support /proc/mm, the host needs to be patched using\n" | ||
294 | " the current skas3 patch.\n\n"); | ||
295 | |||
296 | static int __init noptracefaultinfo_cmd_param(char *str, int* add) | ||
297 | { | ||
298 | ptrace_faultinfo = 0; | ||
299 | return 0; | ||
300 | } | ||
301 | |||
302 | __uml_setup("noptracefaultinfo", noptracefaultinfo_cmd_param, | ||
303 | "noptracefaultinfo\n" | ||
304 | " Turns off usage of PTRACE_FAULTINFO, even if host supports\n" | ||
305 | " it. To support PTRACE_FAULTINFO, the host needs to be patched\n" | ||
306 | " using the current skas3 patch.\n\n"); | ||
307 | |||
308 | #ifdef UML_CONFIG_MODE_SKAS | ||
309 | static inline void check_skas3_ptrace_support(void) | ||
310 | { | ||
311 | struct ptrace_faultinfo fi; | ||
312 | void *stack; | ||
313 | int pid, n; | ||
314 | |||
315 | printf("Checking for the skas3 patch in the host..."); | ||
316 | pid = start_ptraced_child(&stack); | ||
317 | |||
318 | n = ptrace(PTRACE_FAULTINFO, pid, 0, &fi); | ||
319 | if (n < 0) { | ||
320 | ptrace_faultinfo = 0; | ||
321 | if(errno == EIO) | ||
322 | printf("not found\n"); | ||
323 | else | ||
324 | perror("not found"); | ||
325 | } | ||
326 | else { | ||
327 | if (!ptrace_faultinfo) | ||
328 | printf("found but disabled on command line\n"); | ||
329 | else | ||
330 | printf("found\n"); | ||
331 | } | ||
332 | |||
333 | init_registers(pid); | ||
334 | stop_ptraced_child(pid, stack, 1, 1); | ||
335 | } | ||
336 | |||
337 | int can_do_skas(void) | ||
338 | { | ||
339 | printf("Checking for /proc/mm..."); | ||
340 | if (os_access("/proc/mm", OS_ACC_W_OK) < 0) { | ||
341 | proc_mm = 0; | ||
342 | printf("not found\n"); | ||
343 | } | ||
344 | else { | ||
345 | if (!proc_mm) | ||
346 | printf("found but disabled on command line\n"); | ||
347 | else | ||
348 | printf("found\n"); | ||
349 | } | ||
350 | |||
351 | check_skas3_ptrace_support(); | ||
352 | return 1; | ||
353 | } | ||
354 | #else | ||
355 | int can_do_skas(void) | ||
356 | { | ||
357 | return(0); | ||
358 | } | ||
359 | #endif | ||
diff --git a/arch/um/os-Linux/tt.c b/arch/um/os-Linux/tt.c new file mode 100644 index 000000000000..5b047ab8416a --- /dev/null +++ b/arch/um/os-Linux/tt.c | |||
@@ -0,0 +1,113 @@ | |||
1 | /* | ||
2 | * Copyright (C) 2000, 2001, 2002 Jeff Dike (jdike@karaya.com) | ||
3 | * Licensed under the GPL | ||
4 | */ | ||
5 | |||
6 | #include <stdio.h> | ||
7 | #include <unistd.h> | ||
8 | #include <signal.h> | ||
9 | #include <sched.h> | ||
10 | #include <errno.h> | ||
11 | #include <stdarg.h> | ||
12 | #include <stdlib.h> | ||
13 | #include <setjmp.h> | ||
14 | #include <sys/time.h> | ||
15 | #include <sys/ptrace.h> | ||
16 | #include <linux/ptrace.h> | ||
17 | #include <sys/wait.h> | ||
18 | #include <sys/mman.h> | ||
19 | #include <asm/ptrace.h> | ||
20 | #include <asm/unistd.h> | ||
21 | #include <asm/page.h> | ||
22 | #include "user_util.h" | ||
23 | #include "kern_util.h" | ||
24 | #include "user.h" | ||
25 | #include "signal_kern.h" | ||
26 | #include "signal_user.h" | ||
27 | #include "sysdep/ptrace.h" | ||
28 | #include "sysdep/sigcontext.h" | ||
29 | #include "irq_user.h" | ||
30 | #include "ptrace_user.h" | ||
31 | #include "time_user.h" | ||
32 | #include "init.h" | ||
33 | #include "os.h" | ||
34 | #include "uml-config.h" | ||
35 | #include "choose-mode.h" | ||
36 | #include "mode.h" | ||
37 | #include "tempfile.h" | ||
38 | |||
39 | /* | ||
40 | *------------------------- | ||
41 | * only for tt mode (will be deleted in future...) | ||
42 | *------------------------- | ||
43 | */ | ||
44 | |||
45 | struct tramp { | ||
46 | int (*tramp)(void *); | ||
47 | void *tramp_data; | ||
48 | unsigned long temp_stack; | ||
49 | int flags; | ||
50 | int pid; | ||
51 | }; | ||
52 | |||
53 | /* See above for why sigkill is here */ | ||
54 | |||
55 | int sigkill = SIGKILL; | ||
56 | |||
57 | int outer_tramp(void *arg) | ||
58 | { | ||
59 | struct tramp *t; | ||
60 | int sig = sigkill; | ||
61 | |||
62 | t = arg; | ||
63 | t->pid = clone(t->tramp, (void *) t->temp_stack + page_size()/2, | ||
64 | t->flags, t->tramp_data); | ||
65 | if(t->pid > 0) wait_for_stop(t->pid, SIGSTOP, PTRACE_CONT, NULL); | ||
66 | kill(os_getpid(), sig); | ||
67 | _exit(0); | ||
68 | } | ||
69 | |||
70 | int start_fork_tramp(void *thread_arg, unsigned long temp_stack, | ||
71 | int clone_flags, int (*tramp)(void *)) | ||
72 | { | ||
73 | struct tramp arg; | ||
74 | unsigned long sp; | ||
75 | int new_pid, status, err; | ||
76 | |||
77 | /* The trampoline will run on the temporary stack */ | ||
78 | sp = stack_sp(temp_stack); | ||
79 | |||
80 | clone_flags |= CLONE_FILES | SIGCHLD; | ||
81 | |||
82 | arg.tramp = tramp; | ||
83 | arg.tramp_data = thread_arg; | ||
84 | arg.temp_stack = temp_stack; | ||
85 | arg.flags = clone_flags; | ||
86 | |||
87 | /* Start the process and wait for it to kill itself */ | ||
88 | new_pid = clone(outer_tramp, (void *) sp, clone_flags, &arg); | ||
89 | if(new_pid < 0) | ||
90 | return(new_pid); | ||
91 | |||
92 | CATCH_EINTR(err = waitpid(new_pid, &status, 0)); | ||
93 | if(err < 0) | ||
94 | panic("Waiting for outer trampoline failed - errno = %d", | ||
95 | errno); | ||
96 | |||
97 | if(!WIFSIGNALED(status) || (WTERMSIG(status) != SIGKILL)) | ||
98 | panic("outer trampoline didn't exit with SIGKILL, " | ||
99 | "status = %d", status); | ||
100 | |||
101 | return(arg.pid); | ||
102 | } | ||
103 | |||
104 | void forward_pending_sigio(int target) | ||
105 | { | ||
106 | sigset_t sigs; | ||
107 | |||
108 | if(sigpending(&sigs)) | ||
109 | panic("forward_pending_sigio : sigpending failed"); | ||
110 | if(sigismember(&sigs, SIGIO)) | ||
111 | kill(target, SIGIO); | ||
112 | } | ||
113 | |||