xref: /openbmc/qemu/bsd-user/freebsd/os-sys.c (revision 6da777e2)
1 /*
2  *  FreeBSD sysctl() and sysarch() system call emulation
3  *
4  *  Copyright (c) 2013-15 Stacey D. Son
5  *
6  *  This program is free software; you can redistribute it and/or modify
7  *  it under the terms of the GNU General Public License as published by
8  *  the Free Software Foundation; either version 2 of the License, or
9  *  (at your option) any later version.
10  *
11  *  This program is distributed in the hope that it will be useful,
12  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
13  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  *  GNU General Public License for more details.
15  *
16  *  You should have received a copy of the GNU General Public License
17  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
18  */
19 
20 #include "qemu/osdep.h"
21 #include "qemu.h"
22 #include "target_arch_sysarch.h"
23 
24 #include <sys/sysctl.h>
25 
26 /*
27  * Length for the fixed length types.
28  * 0 means variable length for strings and structures
29  * Compare with sys/kern_sysctl.c ctl_size
30  * Note: Not all types appear to be used in-tree.
31  */
32 static const int guest_ctl_size[CTLTYPE + 1] = {
33         [CTLTYPE_INT] = sizeof(abi_int),
34         [CTLTYPE_UINT] = sizeof(abi_uint),
35         [CTLTYPE_LONG] = sizeof(abi_long),
36         [CTLTYPE_ULONG] = sizeof(abi_ulong),
37         [CTLTYPE_S8] = sizeof(int8_t),
38         [CTLTYPE_S16] = sizeof(int16_t),
39         [CTLTYPE_S32] = sizeof(int32_t),
40         [CTLTYPE_S64] = sizeof(int64_t),
41         [CTLTYPE_U8] = sizeof(uint8_t),
42         [CTLTYPE_U16] = sizeof(uint16_t),
43         [CTLTYPE_U32] = sizeof(uint32_t),
44         [CTLTYPE_U64] = sizeof(uint64_t),
45 };
46 
47 static const int host_ctl_size[CTLTYPE + 1] = {
48         [CTLTYPE_INT] = sizeof(int),
49         [CTLTYPE_UINT] = sizeof(u_int),
50         [CTLTYPE_LONG] = sizeof(long),
51         [CTLTYPE_ULONG] = sizeof(u_long),
52         [CTLTYPE_S8] = sizeof(int8_t),
53         [CTLTYPE_S16] = sizeof(int16_t),
54         [CTLTYPE_S32] = sizeof(int32_t),
55         [CTLTYPE_S64] = sizeof(int64_t),
56         [CTLTYPE_U8] = sizeof(uint8_t),
57         [CTLTYPE_U16] = sizeof(uint16_t),
58         [CTLTYPE_U32] = sizeof(uint32_t),
59         [CTLTYPE_U64] = sizeof(uint64_t),
60 };
61 
62 #ifdef TARGET_ABI32
63 /*
64  * Limit the amount of available memory to be most of the 32-bit address
65  * space. 0x100c000 was arrived at through trial and error as a good
66  * definition of 'most'.
67  */
68 static const abi_ulong guest_max_mem = UINT32_MAX - 0x100c000 + 1;
69 
cap_memory(uint64_t mem)70 static abi_ulong cap_memory(uint64_t mem)
71 {
72     return MIN(guest_max_mem, mem);
73 }
74 #endif
75 
scale_to_guest_pages(uint64_t pages)76 static abi_ulong scale_to_guest_pages(uint64_t pages)
77 {
78     /* Scale pages from host to guest */
79     pages = muldiv64(pages, qemu_real_host_page_size(), TARGET_PAGE_SIZE);
80 #ifdef TARGET_ABI32
81     /* cap pages if need be */
82     pages = MIN(pages, guest_max_mem / (abi_ulong)TARGET_PAGE_SIZE);
83 #endif
84     return pages;
85 }
86 
87 #ifdef TARGET_ABI32
88 /* Used only for TARGET_ABI32 */
h2g_long_sat(long l)89 static abi_long h2g_long_sat(long l)
90 {
91     if (l > INT32_MAX) {
92         l = INT32_MAX;
93     } else if (l < INT32_MIN) {
94         l = INT32_MIN;
95     }
96     return l;
97 }
98 
h2g_ulong_sat(u_long ul)99 static abi_ulong h2g_ulong_sat(u_long ul)
100 {
101     return MIN(ul, UINT32_MAX);
102 }
103 #endif
104 
105 /*
106  * placeholder until bsd-user downstream upstreams this with its thread support
107  */
108 #define bsd_get_ncpu() 1
109 
110 /*
111  * This uses the undocumented oidfmt interface to find the kind of a requested
112  * sysctl, see /sys/kern/kern_sysctl.c:sysctl_sysctl_oidfmt() (compare to
113  * src/sbin/sysctl/sysctl.c)
114  */
oidfmt(int * oid,int len,char * fmt,uint32_t * kind)115 static int oidfmt(int *oid, int len, char *fmt, uint32_t *kind)
116 {
117     int qoid[CTL_MAXNAME + 2];
118     uint8_t buf[BUFSIZ];
119     int i;
120     size_t j;
121 
122     qoid[0] = CTL_SYSCTL;
123     qoid[1] = CTL_SYSCTL_OIDFMT;
124     memcpy(qoid + 2, oid, len * sizeof(int));
125 
126     j = sizeof(buf);
127     i = sysctl(qoid, len + 2, buf, &j, 0, 0);
128     if (i) {
129         return i;
130     }
131 
132     if (kind) {
133         *kind = *(uint32_t *)buf;
134     }
135 
136     if (fmt) {
137         strcpy(fmt, (char *)(buf + sizeof(uint32_t)));
138     }
139     return 0;
140 }
141 
142 /*
143  * Convert the old value from host to guest.
144  *
145  * For LONG and ULONG on ABI32, we need to 'down convert' the 8 byte quantities
146  * to 4 bytes. The caller setup a buffer in host memory to get this data from
147  * the kernel and pass it to us. We do the down conversion and adjust the length
148  * so the caller knows what to write as the returned length into the target when
149  * it copies the down converted values into the target.
150  *
151  * For normal integral types, we just need to byte swap. No size changes.
152  *
153  * For strings and node data, there's no conversion needed.
154  *
155  * For opaque data, per sysctl OID converts take care of it.
156  */
h2g_old_sysctl(void * holdp,size_t * holdlen,uint32_t kind)157 static void h2g_old_sysctl(void *holdp, size_t *holdlen, uint32_t kind)
158 {
159     size_t len;
160     int hlen, glen;
161     uint8_t *hp, *gp;
162 
163     /*
164      * Although rare, we can have arrays of sysctl. Both sysctl_old_ddb in
165      * kern_sysctl.c and show_var in sbin/sysctl/sysctl.c have code that loops
166      * this way.  *holdlen has been set by the kernel to the host's length.
167      * Only LONG and ULONG on ABI32 have different sizes: see below.
168      */
169     gp = hp = (uint8_t *)holdp;
170     len = 0;
171     hlen = host_ctl_size[kind & CTLTYPE];
172     glen = guest_ctl_size[kind & CTLTYPE];
173 
174     /*
175      * hlen == 0 for CTLTYPE_STRING and CTLTYPE_NODE, which need no conversion
176      * as well as CTLTYPE_OPAQUE, which needs special converters.
177      */
178     if (hlen == 0) {
179         return;
180     }
181 
182     while (len < *holdlen) {
183         if (hlen == glen) {
184             switch (hlen) {
185             case 1:
186                 /* Nothing needed: no byteswapping and assigning in place */
187                 break;
188             case 2:
189                 *(uint16_t *)gp = tswap16(*(uint16_t *)hp);
190                 break;
191             case 4:
192                 *(uint32_t *)gp = tswap32(*(uint32_t *)hp);
193                 break;
194             case 8:
195                 *(uint64_t *)gp = tswap64(*(uint64_t *)hp);
196                 break;
197             default:
198                 g_assert_not_reached();
199             }
200         } else {
201 #ifdef TARGET_ABI32
202             /*
203              * Saturating assignment for the only two types that differ between
204              * 32-bit and 64-bit machines. All other integral types have the
205              * same, fixed size and will be converted w/o loss of precision
206              * in the above switch.
207              */
208             switch (kind & CTLTYPE) {
209             case CTLTYPE_LONG:
210                 *(abi_long *)gp = tswap32(h2g_long_sat(*(long *)hp));
211                 break;
212             case CTLTYPE_ULONG:
213                 *(abi_ulong *)gp = tswap32(h2g_ulong_sat(*(u_long *)hp));
214                 break;
215             default:
216                 g_assert_not_reached();
217             }
218 #else
219             g_assert_not_reached();
220 #endif
221         }
222         gp += glen;
223         hp += hlen;
224         len += hlen;
225     }
226 #ifdef TARGET_ABI32
227     if (hlen != glen) {
228         *holdlen = (*holdlen / hlen) * glen;
229     }
230 #endif
231 }
232 
233 /*
234  * Convert the undocmented name2oid sysctl data for the target.
235  */
sysctl_name2oid(uint32_t * holdp,size_t holdlen)236 static inline void sysctl_name2oid(uint32_t *holdp, size_t holdlen)
237 {
238     size_t i, num = holdlen / sizeof(uint32_t);
239 
240     for (i = 0; i < num; i++) {
241         holdp[i] = tswap32(holdp[i]);
242     }
243 }
244 
sysctl_oidfmt(uint32_t * holdp)245 static inline void sysctl_oidfmt(uint32_t *holdp)
246 {
247     /* byte swap the kind */
248     holdp[0] = tswap32(holdp[0]);
249 }
250 
do_freebsd_sysctl_oid(CPUArchState * env,int32_t * snamep,int32_t namelen,void * holdp,size_t * holdlenp,void * hnewp,size_t newlen)251 static abi_long do_freebsd_sysctl_oid(CPUArchState *env, int32_t *snamep,
252         int32_t namelen, void *holdp, size_t *holdlenp, void *hnewp,
253         size_t newlen)
254 {
255     uint32_t kind = 0;
256     abi_long ret;
257     size_t holdlen, oldlen;
258 #ifdef TARGET_ABI32
259     void *old_holdp;
260 #endif
261 
262     holdlen = oldlen = *holdlenp;
263     oidfmt(snamep, namelen, NULL, &kind);
264 
265     /* Handle some arch/emulator dependent sysctl()'s here. */
266     switch (snamep[0]) {
267     case CTL_KERN:
268         switch (snamep[1]) {
269         case KERN_USRSTACK:
270             if (oldlen) {
271                 (*(abi_ulong *)holdp) = tswapal(TARGET_USRSTACK);
272             }
273             holdlen = sizeof(abi_ulong);
274             ret = 0;
275             goto out;
276 
277         case KERN_PS_STRINGS:
278             if (oldlen) {
279                 (*(abi_ulong *)holdp) = tswapal(TARGET_PS_STRINGS);
280             }
281             holdlen = sizeof(abi_ulong);
282             ret = 0;
283             goto out;
284 
285         default:
286             break;
287         }
288         break;
289 
290     case CTL_HW:
291         switch (snamep[1]) {
292         case HW_MACHINE:
293             holdlen = sizeof(TARGET_HW_MACHINE);
294             if (holdp) {
295                 strlcpy(holdp, TARGET_HW_MACHINE, oldlen);
296             }
297             ret = 0;
298             goto out;
299 
300         case HW_MACHINE_ARCH:
301         {
302             holdlen = sizeof(TARGET_HW_MACHINE_ARCH);
303             if (holdp) {
304                 strlcpy(holdp, TARGET_HW_MACHINE_ARCH, oldlen);
305             }
306             ret = 0;
307             goto out;
308         }
309         case HW_NCPU:
310             if (oldlen) {
311                 (*(abi_int *)holdp) = tswap32(bsd_get_ncpu());
312             }
313             holdlen = sizeof(int32_t);
314             ret = 0;
315             goto out;
316 #if defined(TARGET_ARM)
317         case HW_FLOATINGPT:
318             if (oldlen) {
319                 ARMCPU *cpu = env_archcpu(env);
320                 *(abi_int *)holdp = cpu_isar_feature(aa32_vfp, cpu);
321             }
322             holdlen = sizeof(abi_int);
323             ret = 0;
324             goto out;
325 #endif
326 
327 
328 #ifdef TARGET_ABI32
329         case HW_PHYSMEM:
330         case HW_USERMEM:
331         case HW_REALMEM:
332             holdlen = sizeof(abi_ulong);
333             ret = 0;
334 
335             if (oldlen) {
336                 int mib[2] = {snamep[0], snamep[1]};
337                 unsigned long lvalue;
338                 size_t len = sizeof(lvalue);
339 
340                 if (sysctl(mib, 2, &lvalue, &len, NULL, 0) == -1) {
341                     ret = -1;
342                 } else {
343                     lvalue = cap_memory(lvalue);
344                     (*(abi_ulong *)holdp) = tswapal((abi_ulong)lvalue);
345                 }
346             }
347             goto out;
348 #endif
349 
350         default:
351         {
352             static int oid_hw_availpages;
353             static int oid_hw_pagesizes;
354 
355             if (!oid_hw_availpages) {
356                 int real_oid[CTL_MAXNAME + 2];
357                 size_t len = sizeof(real_oid) / sizeof(int);
358 
359                 if (sysctlnametomib("hw.availpages", real_oid, &len) >= 0) {
360                     oid_hw_availpages = real_oid[1];
361                 }
362             }
363             if (!oid_hw_pagesizes) {
364                 int real_oid[CTL_MAXNAME + 2];
365                 size_t len = sizeof(real_oid) / sizeof(int);
366 
367                 if (sysctlnametomib("hw.pagesizes", real_oid, &len) >= 0) {
368                     oid_hw_pagesizes = real_oid[1];
369                 }
370             }
371 
372             if (oid_hw_availpages && snamep[1] == oid_hw_availpages) {
373                 long lvalue;
374                 size_t len = sizeof(lvalue);
375 
376                 if (sysctlbyname("hw.availpages", &lvalue, &len, NULL, 0) == -1) {
377                     ret = -1;
378                 } else {
379                     if (oldlen) {
380                         lvalue = scale_to_guest_pages(lvalue);
381                         (*(abi_ulong *)holdp) = tswapal((abi_ulong)lvalue);
382                     }
383                     holdlen = sizeof(abi_ulong);
384                     ret = 0;
385                 }
386                 goto out;
387             }
388 
389             if (oid_hw_pagesizes && snamep[1] == oid_hw_pagesizes) {
390                 if (oldlen) {
391                     (*(abi_ulong *)holdp) = tswapal((abi_ulong)TARGET_PAGE_SIZE);
392                     ((abi_ulong *)holdp)[1] = 0;
393                 }
394                 holdlen = sizeof(abi_ulong) * 2;
395                 ret = 0;
396                 goto out;
397             }
398             break;
399         }
400         }
401         break;
402 
403     default:
404         break;
405     }
406 
407 #ifdef TARGET_ABI32
408     /*
409      * For long and ulong with a 64-bit host and a 32-bit target we have to do
410      * special things. holdlen here is the length provided by the target to the
411      * system call. So we allocate a buffer twice as large because longs are
412      * twice as big on the host which will be writing them. In h2g_old_sysctl
413      * we'll adjust them and adjust the length.
414      */
415     if (kind == CTLTYPE_LONG || kind == CTLTYPE_ULONG) {
416         old_holdp = holdp;
417         holdlen = holdlen * 2;
418         holdp = g_malloc(holdlen);
419     }
420 #endif
421 
422     ret = get_errno(sysctl(snamep, namelen, holdp, &holdlen, hnewp, newlen));
423     if (!ret && (holdp != 0)) {
424 
425         if (snamep[0] == CTL_SYSCTL) {
426             switch (snamep[1]) {
427             case CTL_SYSCTL_NEXT:
428             case CTL_SYSCTL_NAME2OID:
429             case CTL_SYSCTL_NEXTNOSKIP:
430                 /*
431                  * All of these return an OID array, so we need to convert to
432                  * target.
433                  */
434                 sysctl_name2oid(holdp, holdlen);
435                 break;
436 
437             case CTL_SYSCTL_OIDFMT:
438                 /* Handle oidfmt */
439                 sysctl_oidfmt(holdp);
440                 break;
441             case CTL_SYSCTL_OIDDESCR:
442             case CTL_SYSCTL_OIDLABEL:
443             default:
444                 /* Handle it based on the type */
445                 h2g_old_sysctl(holdp, &holdlen, kind);
446                 /* NB: None of these are LONG or ULONG */
447                 break;
448             }
449         } else {
450             /*
451              * Need to convert from host to target. All the weird special cases
452              * are handled above.
453              */
454             h2g_old_sysctl(holdp, &holdlen, kind);
455 #ifdef TARGET_ABI32
456             /*
457              * For the 32-bit on 64-bit case, for longs we need to copy the
458              * now-converted buffer to the target and free the buffer.
459              */
460             if (kind == CTLTYPE_LONG || kind == CTLTYPE_ULONG) {
461                 memcpy(old_holdp, holdp, holdlen);
462                 g_free(holdp);
463                 holdp = old_holdp;
464             }
465 #endif
466         }
467     }
468 
469 out:
470     *holdlenp = holdlen;
471     return ret;
472 }
473 
474 /*
475  * This syscall was created to make sysctlbyname(3) more efficient, but we can't
476  * really provide it in bsd-user.  Notably, we must always translate the names
477  * independently since some sysctl values have to be faked for the target
478  * environment, so it still has to break down to two syscalls for the underlying
479  * implementation.
480  */
do_freebsd_sysctlbyname(CPUArchState * env,abi_ulong namep,int32_t namelen,abi_ulong oldp,abi_ulong oldlenp,abi_ulong newp,abi_ulong newlen)481 abi_long do_freebsd_sysctlbyname(CPUArchState *env, abi_ulong namep,
482         int32_t namelen, abi_ulong oldp, abi_ulong oldlenp, abi_ulong newp,
483         abi_ulong newlen)
484 {
485     abi_long ret = -TARGET_EFAULT;
486     void *holdp = NULL, *hnewp = NULL;
487     char *snamep = NULL;
488     int oid[CTL_MAXNAME + 2];
489     size_t holdlen, oidplen;
490     abi_ulong oldlen = 0;
491 
492     /* oldlenp is read/write, pre-check here for write */
493     if (oldlenp) {
494         if (!access_ok(VERIFY_WRITE, oldlenp, sizeof(abi_ulong)) ||
495             get_user_ual(oldlen, oldlenp)) {
496             goto out;
497         }
498     }
499     snamep = lock_user_string(namep);
500     if (snamep == NULL) {
501         goto out;
502     }
503     if (newp) {
504         hnewp = lock_user(VERIFY_READ, newp, newlen, 1);
505         if (hnewp == NULL) {
506             goto out;
507         }
508     }
509     if (oldp) {
510         holdp = lock_user(VERIFY_WRITE, oldp, oldlen, 0);
511         if (holdp == NULL) {
512             goto out;
513         }
514     }
515     holdlen = oldlen;
516 
517     oidplen = ARRAY_SIZE(oid);
518     if (sysctlnametomib(snamep, oid, &oidplen) != 0) {
519         ret = -TARGET_EINVAL;
520         goto out;
521     }
522 
523     ret = do_freebsd_sysctl_oid(env, oid, oidplen, holdp, &holdlen, hnewp,
524         newlen);
525 
526     /*
527      * writeability pre-checked above. __sysctl(2) returns ENOMEM and updates
528      * oldlenp for the proper size to use.
529      */
530     if (oldlenp && (ret == 0 || ret == -TARGET_ENOMEM)) {
531         put_user_ual(holdlen, oldlenp);
532     }
533 out:
534     unlock_user(snamep, namep, 0);
535     unlock_user(holdp, oldp, ret == 0 ? holdlen : 0);
536     unlock_user(hnewp, newp, 0);
537 
538     return ret;
539 }
540 
do_freebsd_sysctl(CPUArchState * env,abi_ulong namep,int32_t namelen,abi_ulong oldp,abi_ulong oldlenp,abi_ulong newp,abi_ulong newlen)541 abi_long do_freebsd_sysctl(CPUArchState *env, abi_ulong namep, int32_t namelen,
542         abi_ulong oldp, abi_ulong oldlenp, abi_ulong newp, abi_ulong newlen)
543 {
544     abi_long ret = -TARGET_EFAULT;
545     void *hnamep, *holdp = NULL, *hnewp = NULL;
546     size_t holdlen;
547     abi_ulong oldlen = 0;
548     int32_t *snamep = g_malloc(sizeof(int32_t) * namelen), *p, *q, i;
549 
550     /* oldlenp is read/write, pre-check here for write */
551     if (oldlenp) {
552         if (!access_ok(VERIFY_WRITE, oldlenp, sizeof(abi_ulong)) ||
553             get_user_ual(oldlen, oldlenp)) {
554             goto out;
555         }
556     }
557     hnamep = lock_user(VERIFY_READ, namep, namelen, 1);
558     if (hnamep == NULL) {
559         goto out;
560     }
561     if (newp) {
562         hnewp = lock_user(VERIFY_READ, newp, newlen, 1);
563         if (hnewp == NULL) {
564             goto out;
565         }
566     }
567     if (oldp) {
568         holdp = lock_user(VERIFY_WRITE, oldp, oldlen, 0);
569         if (holdp == NULL) {
570             goto out;
571         }
572     }
573     holdlen = oldlen;
574     for (p = hnamep, q = snamep, i = 0; i < namelen; p++, i++, q++) {
575         *q = tswap32(*p);
576     }
577 
578     ret = do_freebsd_sysctl_oid(env, snamep, namelen, holdp, &holdlen, hnewp,
579         newlen);
580 
581     /*
582      * writeability pre-checked above. __sysctl(2) returns ENOMEM and updates
583      * oldlenp for the proper size to use.
584      */
585     if (oldlenp && (ret == 0 || ret == -TARGET_ENOMEM)) {
586         put_user_ual(holdlen, oldlenp);
587     }
588     unlock_user(hnamep, namep, 0);
589     unlock_user(holdp, oldp, ret == 0 ? holdlen : 0);
590 out:
591     g_free(snamep);
592     return ret;
593 }
594 
595 /* sysarch() is architecture dependent. */
do_freebsd_sysarch(void * cpu_env,abi_long arg1,abi_long arg2)596 abi_long do_freebsd_sysarch(void *cpu_env, abi_long arg1, abi_long arg2)
597 {
598     return do_freebsd_arch_sysarch(cpu_env, arg1, arg2);
599 }
600