The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/i386/i386/sys_machdep.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (c) 1990 The Regents of the University of California.
    5  * All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. Neither the name of the University nor the names of its contributors
   16  *    may be used to endorse or promote products derived from this software
   17  *    without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  *
   31  *      from: @(#)sys_machdep.c 5.5 (Berkeley) 1/19/91
   32  */
   33 
   34 #include <sys/cdefs.h>
   35 __FBSDID("$FreeBSD$");
   36 
   37 #include "opt_capsicum.h"
   38 #include "opt_kstack_pages.h"
   39 
   40 #include <sys/param.h>
   41 #include <sys/capsicum.h>
   42 #include <sys/systm.h>
   43 #include <sys/lock.h>
   44 #include <sys/malloc.h>
   45 #include <sys/mutex.h>
   46 #include <sys/priv.h>
   47 #include <sys/proc.h>
   48 #include <sys/smp.h>
   49 #include <sys/sysproto.h>
   50 
   51 #include <vm/vm.h>
   52 #include <vm/pmap.h>
   53 #include <vm/vm_map.h>
   54 #include <vm/vm_extern.h>
   55 
   56 #include <machine/atomic.h>
   57 #include <machine/cpu.h>
   58 #include <machine/pcb.h>
   59 #include <machine/pcb_ext.h>
   60 #include <machine/proc.h>
   61 #include <machine/sysarch.h>
   62 
   63 #include <security/audit/audit.h>
   64 
   65 #include <vm/vm_kern.h>         /* for kernel_map */
   66 
   67 #define MAX_LD 8192
   68 #define LD_PER_PAGE 512
   69 #define NEW_MAX_LD(num)  rounddown2(num + LD_PER_PAGE, LD_PER_PAGE)
   70 #define SIZE_FROM_LARGEST_LD(num) (NEW_MAX_LD(num) << 3)
   71 #define NULL_LDT_BASE   ((caddr_t)NULL)
   72 
   73 #ifdef SMP
   74 static void set_user_ldt_rv(void *arg);
   75 #endif
   76 static int i386_set_ldt_data(struct thread *, int start, int num,
   77     union descriptor *descs);
   78 static int i386_ldt_grow(struct thread *td, int len);
   79 
   80 void
   81 fill_based_sd(struct segment_descriptor *sdp, uint32_t base)
   82 {
   83 
   84         sdp->sd_lobase = base & 0xffffff;
   85         sdp->sd_hibase = (base >> 24) & 0xff;
   86         sdp->sd_lolimit = 0xffff;       /* 4GB limit, wraps around */
   87         sdp->sd_hilimit = 0xf;
   88         sdp->sd_type = SDT_MEMRWA;
   89         sdp->sd_dpl = SEL_UPL;
   90         sdp->sd_p = 1;
   91         sdp->sd_xx = 0;
   92         sdp->sd_def32 = 1;
   93         sdp->sd_gran = 1;
   94 }
   95 
   96 /*
   97  * Construct special descriptors for "base" selectors.  Store them in
   98  * the PCB for later use by cpu_switch().  Store them in the GDT for
   99  * more immediate use.  The GDT entries are part of the current
  100  * context.  Callers must load related segment registers to complete
  101  * setting up the current context.
  102  */
  103 void
  104 set_fsbase(struct thread *td, uint32_t base)
  105 {
  106         struct segment_descriptor sd;
  107 
  108         fill_based_sd(&sd, base);
  109         critical_enter();
  110         td->td_pcb->pcb_fsd = sd;
  111         if (td == curthread)
  112                 PCPU_GET(fsgs_gdt)[0] = sd;
  113         critical_exit();
  114 }
  115 
  116 void
  117 set_gsbase(struct thread *td, uint32_t base)
  118 {
  119         struct segment_descriptor sd;
  120 
  121         fill_based_sd(&sd, base);
  122         critical_enter();
  123         td->td_pcb->pcb_gsd = sd;
  124         if (td == curthread)
  125                 PCPU_GET(fsgs_gdt)[1] = sd;
  126         critical_exit();
  127 }
  128 
  129 #ifndef _SYS_SYSPROTO_H_
  130 struct sysarch_args {
  131         int op;
  132         char *parms;
  133 };
  134 #endif
  135 
  136 int
  137 sysarch(struct thread *td, struct sysarch_args *uap)
  138 {
  139         int error;
  140         union descriptor *lp;
  141         union {
  142                 struct i386_ldt_args largs;
  143                 struct i386_ioperm_args iargs;
  144                 struct i386_get_xfpustate xfpu;
  145         } kargs;
  146         uint32_t base;
  147         struct segment_descriptor *sdp;
  148 
  149         AUDIT_ARG_CMD(uap->op);
  150 
  151 #ifdef CAPABILITY_MODE
  152         /*
  153          * When adding new operations, add a new case statement here to
  154          * explicitly indicate whether or not the operation is safe to
  155          * perform in capability mode.
  156          */
  157         if (IN_CAPABILITY_MODE(td)) {
  158                 switch (uap->op) {
  159                 case I386_GET_LDT:
  160                 case I386_SET_LDT:
  161                 case I386_GET_IOPERM:
  162                 case I386_GET_FSBASE:
  163                 case I386_SET_FSBASE:
  164                 case I386_GET_GSBASE:
  165                 case I386_SET_GSBASE:
  166                 case I386_GET_XFPUSTATE:
  167                         break;
  168 
  169                 case I386_SET_IOPERM:
  170                 default:
  171 #ifdef KTRACE
  172                         if (KTRPOINT(td, KTR_CAPFAIL))
  173                                 ktrcapfail(CAPFAIL_SYSCALL, NULL, NULL);
  174 #endif
  175                         return (ECAPMODE);
  176                 }
  177         }
  178 #endif
  179 
  180         switch (uap->op) {
  181         case I386_GET_IOPERM:
  182         case I386_SET_IOPERM:
  183                 if ((error = copyin(uap->parms, &kargs.iargs,
  184                     sizeof(struct i386_ioperm_args))) != 0)
  185                         return (error);
  186                 break;
  187         case I386_GET_LDT:
  188         case I386_SET_LDT:
  189                 if ((error = copyin(uap->parms, &kargs.largs,
  190                     sizeof(struct i386_ldt_args))) != 0)
  191                         return (error);
  192                 break;
  193         case I386_GET_XFPUSTATE:
  194                 if ((error = copyin(uap->parms, &kargs.xfpu,
  195                     sizeof(struct i386_get_xfpustate))) != 0)
  196                         return (error);
  197                 break;
  198         default:
  199                 break;
  200         }
  201 
  202         switch (uap->op) {
  203         case I386_GET_LDT:
  204                 error = i386_get_ldt(td, &kargs.largs);
  205                 break;
  206         case I386_SET_LDT:
  207                 if (kargs.largs.descs != NULL) {
  208                         if (kargs.largs.num > MAX_LD)
  209                                 return (EINVAL);
  210                         lp = malloc(kargs.largs.num * sizeof(union descriptor),
  211                             M_TEMP, M_WAITOK);
  212                         error = copyin(kargs.largs.descs, lp,
  213                             kargs.largs.num * sizeof(union descriptor));
  214                         if (error == 0)
  215                                 error = i386_set_ldt(td, &kargs.largs, lp);
  216                         free(lp, M_TEMP);
  217                 } else {
  218                         error = i386_set_ldt(td, &kargs.largs, NULL);
  219                 }
  220                 break;
  221         case I386_GET_IOPERM:
  222                 error = i386_get_ioperm(td, &kargs.iargs);
  223                 if (error == 0)
  224                         error = copyout(&kargs.iargs, uap->parms,
  225                             sizeof(struct i386_ioperm_args));
  226                 break;
  227         case I386_SET_IOPERM:
  228                 error = i386_set_ioperm(td, &kargs.iargs);
  229                 break;
  230         case I386_VM86:
  231                 error = vm86_sysarch(td, uap->parms);
  232                 break;
  233         case I386_GET_FSBASE:
  234                 sdp = &td->td_pcb->pcb_fsd;
  235                 base = sdp->sd_hibase << 24 | sdp->sd_lobase;
  236                 error = copyout(&base, uap->parms, sizeof(base));
  237                 break;
  238         case I386_SET_FSBASE:
  239                 error = copyin(uap->parms, &base, sizeof(base));
  240                 if (error == 0) {
  241                         /*
  242                          * Construct the special descriptor for fsbase
  243                          * and arrange for doreti to load its selector
  244                          * soon enough.
  245                          */
  246                         set_fsbase(td, base);
  247                         td->td_frame->tf_fs = GSEL(GUFS_SEL, SEL_UPL);
  248                 }
  249                 break;
  250         case I386_GET_GSBASE:
  251                 sdp = &td->td_pcb->pcb_gsd;
  252                 base = sdp->sd_hibase << 24 | sdp->sd_lobase;
  253                 error = copyout(&base, uap->parms, sizeof(base));
  254                 break;
  255         case I386_SET_GSBASE:
  256                 error = copyin(uap->parms, &base, sizeof(base));
  257                 if (error == 0) {
  258                         /*
  259                          * Construct the special descriptor for gsbase.
  260                          * The selector is loaded immediately, since we
  261                          * normally only reload %gs on context switches.
  262                          */
  263                         set_gsbase(td, base);
  264                         load_gs(GSEL(GUGS_SEL, SEL_UPL));
  265                 }
  266                 break;
  267         case I386_GET_XFPUSTATE:
  268                 if (kargs.xfpu.len > cpu_max_ext_state_size -
  269                     sizeof(union savefpu))
  270                         return (EINVAL);
  271                 npxgetregs(td);
  272                 error = copyout((char *)(get_pcb_user_save_td(td) + 1),
  273                     kargs.xfpu.addr, kargs.xfpu.len);
  274                 break;
  275         default:
  276                 error = EINVAL;
  277                 break;
  278         }
  279         return (error);
  280 }
  281 
  282 int
  283 i386_extend_pcb(struct thread *td)
  284 {
  285         int i, offset;
  286         u_long *addr;
  287         struct pcb_ext *ext;
  288         struct soft_segment_descriptor ssd = {
  289                 0,                      /* segment base address (overwritten) */
  290                 ctob(IOPAGES + 1) - 1,  /* length */
  291                 SDT_SYS386TSS,          /* segment type */
  292                 0,                      /* priority level */
  293                 1,                      /* descriptor present */
  294                 0, 0,
  295                 0,                      /* default 32 size */
  296                 0                       /* granularity */
  297         };
  298 
  299         ext = pmap_trm_alloc(ctob(IOPAGES + 1), M_WAITOK | M_ZERO);
  300         /* -16 is so we can convert a trapframe into vm86trapframe inplace */
  301         ext->ext_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
  302         /*
  303          * The last byte of the i/o map must be followed by an 0xff byte.
  304          * We arbitrarily allocate 16 bytes here, to keep the starting
  305          * address on a doubleword boundary.
  306          */
  307         offset = PAGE_SIZE - 16;
  308         ext->ext_tss.tss_ioopt = 
  309             (offset - ((unsigned)&ext->ext_tss - (unsigned)ext)) << 16;
  310         ext->ext_iomap = (caddr_t)ext + offset;
  311         ext->ext_vm86.vm86_intmap = (caddr_t)ext + offset - 32;
  312 
  313         addr = (u_long *)ext->ext_vm86.vm86_intmap;
  314         for (i = 0; i < (ctob(IOPAGES) + 32 + 16) / sizeof(u_long); i++)
  315                 *addr++ = ~0;
  316 
  317         ssd.ssd_base = (unsigned)&ext->ext_tss;
  318         ssd.ssd_limit -= ((unsigned)&ext->ext_tss - (unsigned)ext);
  319         ssdtosd(&ssd, &ext->ext_tssd);
  320 
  321         KASSERT(td == curthread, ("giving TSS to !curthread"));
  322         KASSERT(td->td_pcb->pcb_ext == 0, ("already have a TSS!"));
  323 
  324         /* Switch to the new TSS. */
  325         critical_enter();
  326         ext->ext_tss.tss_esp0 = PCPU_GET(trampstk);
  327         td->td_pcb->pcb_ext = ext;
  328         PCPU_SET(private_tss, 1);
  329         *PCPU_GET(tss_gdt) = ext->ext_tssd;
  330         ltr(GSEL(GPROC0_SEL, SEL_KPL));
  331         critical_exit();
  332 
  333         return 0;
  334 }
  335 
  336 int
  337 i386_set_ioperm(td, uap)
  338         struct thread *td;
  339         struct i386_ioperm_args *uap;
  340 {
  341         char *iomap;
  342         u_int i;
  343         int error;
  344 
  345         if ((error = priv_check(td, PRIV_IO)) != 0)
  346                 return (error);
  347         if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
  348                 return (error);
  349         /*
  350          * XXX 
  351          * While this is restricted to root, we should probably figure out
  352          * whether any other driver is using this i/o address, as so not to
  353          * cause confusion.  This probably requires a global 'usage registry'.
  354          */
  355 
  356         if (td->td_pcb->pcb_ext == 0)
  357                 if ((error = i386_extend_pcb(td)) != 0)
  358                         return (error);
  359         iomap = (char *)td->td_pcb->pcb_ext->ext_iomap;
  360 
  361         if (uap->start > uap->start + uap->length ||
  362             uap->start + uap->length > IOPAGES * PAGE_SIZE * NBBY)
  363                 return (EINVAL);
  364 
  365         for (i = uap->start; i < uap->start + uap->length; i++) {
  366                 if (uap->enable)
  367                         iomap[i >> 3] &= ~(1 << (i & 7));
  368                 else
  369                         iomap[i >> 3] |= (1 << (i & 7));
  370         }
  371         return (error);
  372 }
  373 
  374 int
  375 i386_get_ioperm(td, uap)
  376         struct thread *td;
  377         struct i386_ioperm_args *uap;
  378 {
  379         int i, state;
  380         char *iomap;
  381 
  382         if (uap->start >= IOPAGES * PAGE_SIZE * NBBY)
  383                 return (EINVAL);
  384 
  385         if (td->td_pcb->pcb_ext == 0) {
  386                 uap->length = 0;
  387                 goto done;
  388         }
  389 
  390         iomap = (char *)td->td_pcb->pcb_ext->ext_iomap;
  391 
  392         i = uap->start;
  393         state = (iomap[i >> 3] >> (i & 7)) & 1;
  394         uap->enable = !state;
  395         uap->length = 1;
  396 
  397         for (i = uap->start + 1; i < IOPAGES * PAGE_SIZE * NBBY; i++) {
  398                 if (state != ((iomap[i >> 3] >> (i & 7)) & 1))
  399                         break;
  400                 uap->length++;
  401         }
  402 
  403 done:
  404         return (0);
  405 }
  406 
  407 /*
  408  * Update the GDT entry pointing to the LDT to point to the LDT of the
  409  * current process. Manage dt_lock holding/unholding autonomously.
  410  */   
  411 static void
  412 set_user_ldt_locked(struct mdproc *mdp)
  413 {
  414         struct proc_ldt *pldt;
  415         int gdt_idx;
  416 
  417         mtx_assert(&dt_lock, MA_OWNED);
  418 
  419         pldt = mdp->md_ldt;
  420         gdt_idx = GUSERLDT_SEL;
  421         gdt_idx += PCPU_GET(cpuid) * NGDT;      /* always 0 on UP */
  422         gdt[gdt_idx].sd = pldt->ldt_sd;
  423         lldt(GSEL(GUSERLDT_SEL, SEL_KPL));
  424         PCPU_SET(currentldt, GSEL(GUSERLDT_SEL, SEL_KPL));
  425 }
  426 
  427 void
  428 set_user_ldt(struct mdproc *mdp)
  429 {
  430 
  431         mtx_lock_spin(&dt_lock);
  432         set_user_ldt_locked(mdp);
  433         mtx_unlock_spin(&dt_lock);
  434 }
  435 
  436 #ifdef SMP
  437 static void
  438 set_user_ldt_rv(void *arg)
  439 {
  440         struct proc *p;
  441 
  442         p = curproc;
  443         if (arg == p->p_vmspace)
  444                 set_user_ldt(&p->p_md);
  445 }
  446 #endif
  447 
  448 /*
  449  * dt_lock must be held. Returns with dt_lock held.
  450  */
  451 struct proc_ldt *
  452 user_ldt_alloc(struct mdproc *mdp, int len)
  453 {
  454         struct proc_ldt *pldt, *new_ldt;
  455 
  456         mtx_assert(&dt_lock, MA_OWNED);
  457         mtx_unlock_spin(&dt_lock);
  458         new_ldt = malloc(sizeof(struct proc_ldt), M_SUBPROC, M_WAITOK);
  459 
  460         new_ldt->ldt_len = len = NEW_MAX_LD(len);
  461         new_ldt->ldt_base = pmap_trm_alloc(len * sizeof(union descriptor),
  462             M_WAITOK | M_ZERO);
  463         new_ldt->ldt_refcnt = 1;
  464         new_ldt->ldt_active = 0;
  465 
  466         mtx_lock_spin(&dt_lock);
  467         gdt_segs[GUSERLDT_SEL].ssd_base = (unsigned)new_ldt->ldt_base;
  468         gdt_segs[GUSERLDT_SEL].ssd_limit = len * sizeof(union descriptor) - 1;
  469         ssdtosd(&gdt_segs[GUSERLDT_SEL], &new_ldt->ldt_sd);
  470 
  471         if ((pldt = mdp->md_ldt) != NULL) {
  472                 if (len > pldt->ldt_len)
  473                         len = pldt->ldt_len;
  474                 bcopy(pldt->ldt_base, new_ldt->ldt_base,
  475                     len * sizeof(union descriptor));
  476         } else
  477                 bcopy(ldt, new_ldt->ldt_base, sizeof(union descriptor) * NLDT);
  478 
  479         return (new_ldt);
  480 }
  481 
  482 /*
  483  * Must be called with dt_lock held.  Returns with dt_lock unheld.
  484  */
  485 void
  486 user_ldt_free(struct thread *td)
  487 {
  488         struct mdproc *mdp;
  489         struct proc_ldt *pldt;
  490 
  491         mtx_assert(&dt_lock, MA_OWNED);
  492         mdp = &td->td_proc->p_md;
  493         if ((pldt = mdp->md_ldt) == NULL) {
  494                 mtx_unlock_spin(&dt_lock);
  495                 return;
  496         }
  497 
  498         if (td == curthread) {
  499                 lldt(_default_ldt);
  500                 PCPU_SET(currentldt, _default_ldt);
  501         }
  502 
  503         mdp->md_ldt = NULL;
  504         user_ldt_deref(pldt);
  505 }
  506 
  507 void
  508 user_ldt_deref(struct proc_ldt *pldt)
  509 {
  510 
  511         mtx_assert(&dt_lock, MA_OWNED);
  512         if (--pldt->ldt_refcnt == 0) {
  513                 mtx_unlock_spin(&dt_lock);
  514                 pmap_trm_free(pldt->ldt_base, pldt->ldt_len *
  515                     sizeof(union descriptor));
  516                 free(pldt, M_SUBPROC);
  517         } else
  518                 mtx_unlock_spin(&dt_lock);
  519 }
  520 
  521 /*
  522  * Note for the authors of compat layers (linux, etc): copyout() in
  523  * the function below is not a problem since it presents data in
  524  * arch-specific format (i.e. i386-specific in this case), not in
  525  * the OS-specific one.
  526  */
  527 int
  528 i386_get_ldt(struct thread *td, struct i386_ldt_args *uap)
  529 {
  530         struct proc_ldt *pldt;
  531         char *data;
  532         u_int nldt, num;
  533         int error;
  534 
  535 #ifdef DEBUG
  536         printf("i386_get_ldt: start=%u num=%u descs=%p\n",
  537             uap->start, uap->num, (void *)uap->descs);
  538 #endif
  539 
  540         num = min(uap->num, MAX_LD);
  541         data = malloc(num * sizeof(union descriptor), M_TEMP, M_WAITOK);
  542         mtx_lock_spin(&dt_lock);
  543         pldt = td->td_proc->p_md.md_ldt;
  544         nldt = pldt != NULL ? pldt->ldt_len : NLDT;
  545         if (uap->start >= nldt) {
  546                 num = 0;
  547         } else {
  548                 num = min(num, nldt - uap->start);
  549                 bcopy(pldt != NULL ?
  550                     &((union descriptor *)(pldt->ldt_base))[uap->start] :
  551                     &ldt[uap->start], data, num * sizeof(union descriptor));
  552         }
  553         mtx_unlock_spin(&dt_lock);
  554         error = copyout(data, uap->descs, num * sizeof(union descriptor));
  555         if (error == 0)
  556                 td->td_retval[0] = num;
  557         free(data, M_TEMP);
  558         return (error);
  559 }
  560 
  561 int
  562 i386_set_ldt(struct thread *td, struct i386_ldt_args *uap,
  563     union descriptor *descs)
  564 {
  565         struct mdproc *mdp;
  566         struct proc_ldt *pldt;
  567         union descriptor *dp;
  568         u_int largest_ld, i;
  569         int error;
  570 
  571 #ifdef DEBUG
  572         printf("i386_set_ldt: start=%u num=%u descs=%p\n",
  573             uap->start, uap->num, (void *)uap->descs);
  574 #endif
  575         error = 0;
  576         mdp = &td->td_proc->p_md;
  577 
  578         if (descs == NULL) {
  579                 /* Free descriptors */
  580                 if (uap->start == 0 && uap->num == 0) {
  581                         /*
  582                          * Treat this as a special case, so userland needn't
  583                          * know magic number NLDT.
  584                          */
  585                         uap->start = NLDT;
  586                         uap->num = MAX_LD - NLDT;
  587                 }
  588                 mtx_lock_spin(&dt_lock);
  589                 if ((pldt = mdp->md_ldt) == NULL ||
  590                     uap->start >= pldt->ldt_len) {
  591                         mtx_unlock_spin(&dt_lock);
  592                         return (0);
  593                 }
  594                 largest_ld = uap->start + uap->num;
  595                 if (largest_ld > pldt->ldt_len)
  596                         largest_ld = pldt->ldt_len;
  597                 for (i = uap->start; i < largest_ld; i++)
  598                         atomic_store_rel_64(&((uint64_t *)(pldt->ldt_base))[i],
  599                             0);
  600                 mtx_unlock_spin(&dt_lock);
  601                 return (0);
  602         }
  603 
  604         if (uap->start != LDT_AUTO_ALLOC || uap->num != 1) {
  605                 /* verify range of descriptors to modify */
  606                 largest_ld = uap->start + uap->num;
  607                 if (uap->start >= MAX_LD || largest_ld > MAX_LD)
  608                         return (EINVAL);
  609         }
  610 
  611         /* Check descriptors for access violations */
  612         for (i = 0; i < uap->num; i++) {
  613                 dp = &descs[i];
  614 
  615                 switch (dp->sd.sd_type) {
  616                 case SDT_SYSNULL:       /* system null */ 
  617                         dp->sd.sd_p = 0;
  618                         break;
  619                 case SDT_SYS286TSS: /* system 286 TSS available */
  620                 case SDT_SYSLDT:    /* system local descriptor table */
  621                 case SDT_SYS286BSY: /* system 286 TSS busy */
  622                 case SDT_SYSTASKGT: /* system task gate */
  623                 case SDT_SYS286IGT: /* system 286 interrupt gate */
  624                 case SDT_SYS286TGT: /* system 286 trap gate */
  625                 case SDT_SYSNULL2:  /* undefined by Intel */ 
  626                 case SDT_SYS386TSS: /* system 386 TSS available */
  627                 case SDT_SYSNULL3:  /* undefined by Intel */
  628                 case SDT_SYS386BSY: /* system 386 TSS busy */
  629                 case SDT_SYSNULL4:  /* undefined by Intel */ 
  630                 case SDT_SYS386IGT: /* system 386 interrupt gate */
  631                 case SDT_SYS386TGT: /* system 386 trap gate */
  632                 case SDT_SYS286CGT: /* system 286 call gate */ 
  633                 case SDT_SYS386CGT: /* system 386 call gate */
  634                         return (EACCES);
  635 
  636                 /* memory segment types */
  637                 case SDT_MEMEC:   /* memory execute only conforming */
  638                 case SDT_MEMEAC:  /* memory execute only accessed conforming */
  639                 case SDT_MEMERC:  /* memory execute read conforming */
  640                 case SDT_MEMERAC: /* memory execute read accessed conforming */
  641                          /* Must be "present" if executable and conforming. */
  642                         if (dp->sd.sd_p == 0)
  643                                 return (EACCES);
  644                         break;
  645                 case SDT_MEMRO:   /* memory read only */
  646                 case SDT_MEMROA:  /* memory read only accessed */
  647                 case SDT_MEMRW:   /* memory read write */
  648                 case SDT_MEMRWA:  /* memory read write accessed */
  649                 case SDT_MEMROD:  /* memory read only expand dwn limit */
  650                 case SDT_MEMRODA: /* memory read only expand dwn lim accessed */
  651                 case SDT_MEMRWD:  /* memory read write expand dwn limit */  
  652                 case SDT_MEMRWDA: /* memory read write expand dwn lim acessed */
  653                 case SDT_MEME:    /* memory execute only */ 
  654                 case SDT_MEMEA:   /* memory execute only accessed */
  655                 case SDT_MEMER:   /* memory execute read */
  656                 case SDT_MEMERA:  /* memory execute read accessed */
  657                         break;
  658                 default:
  659                         return (EINVAL);
  660                 }
  661 
  662                 /* Only user (ring-3) descriptors may be present. */
  663                 if (dp->sd.sd_p != 0 && dp->sd.sd_dpl != SEL_UPL)
  664                         return (EACCES);
  665         }
  666 
  667         if (uap->start == LDT_AUTO_ALLOC && uap->num == 1) {
  668                 /* Allocate a free slot */
  669                 mtx_lock_spin(&dt_lock);
  670                 if ((pldt = mdp->md_ldt) == NULL) {
  671                         if ((error = i386_ldt_grow(td, NLDT + 1))) {
  672                                 mtx_unlock_spin(&dt_lock);
  673                                 return (error);
  674                         }
  675                         pldt = mdp->md_ldt;
  676                 }
  677 again:
  678                 /*
  679                  * start scanning a bit up to leave room for NVidia and
  680                  * Wine, which still user the "Blat" method of allocation.
  681                  */
  682                 dp = &((union descriptor *)(pldt->ldt_base))[NLDT];
  683                 for (i = NLDT; i < pldt->ldt_len; ++i) {
  684                         if (dp->sd.sd_type == SDT_SYSNULL)
  685                                 break;
  686                         dp++;
  687                 }
  688                 if (i >= pldt->ldt_len) {
  689                         if ((error = i386_ldt_grow(td, pldt->ldt_len+1))) {
  690                                 mtx_unlock_spin(&dt_lock);
  691                                 return (error);
  692                         }
  693                         goto again;
  694                 }
  695                 uap->start = i;
  696                 error = i386_set_ldt_data(td, i, 1, descs);
  697                 mtx_unlock_spin(&dt_lock);
  698         } else {
  699                 largest_ld = uap->start + uap->num;
  700                 mtx_lock_spin(&dt_lock);
  701                 if (!(error = i386_ldt_grow(td, largest_ld))) {
  702                         error = i386_set_ldt_data(td, uap->start, uap->num,
  703                             descs);
  704                 }
  705                 mtx_unlock_spin(&dt_lock);
  706         }
  707         if (error == 0)
  708                 td->td_retval[0] = uap->start;
  709         return (error);
  710 }
  711 
  712 static int
  713 i386_set_ldt_data(struct thread *td, int start, int num,
  714     union descriptor *descs)
  715 {
  716         struct mdproc *mdp;
  717         struct proc_ldt *pldt;
  718         uint64_t *dst, *src;
  719         int i;
  720 
  721         mtx_assert(&dt_lock, MA_OWNED);
  722 
  723         mdp = &td->td_proc->p_md;
  724         pldt = mdp->md_ldt;
  725         dst = (uint64_t *)(pldt->ldt_base);
  726         src = (uint64_t *)descs;
  727 
  728         /*
  729          * Atomic(9) is used only to get 64bit atomic store with
  730          * cmpxchg8b when available.  There is no op without release
  731          * semantic.
  732          */
  733         for (i = 0; i < num; i++)
  734                 atomic_store_rel_64(&dst[start + i], src[i]);
  735         return (0);
  736 }
  737 
  738 static int
  739 i386_ldt_grow(struct thread *td, int len) 
  740 {
  741         struct mdproc *mdp;
  742         struct proc_ldt *new_ldt, *pldt;
  743         caddr_t old_ldt_base;
  744         int old_ldt_len;
  745 
  746         mtx_assert(&dt_lock, MA_OWNED);
  747 
  748         if (len > MAX_LD)
  749                 return (ENOMEM);
  750         if (len < NLDT + 1)
  751                 len = NLDT + 1;
  752 
  753         mdp = &td->td_proc->p_md;
  754         old_ldt_base = NULL_LDT_BASE;
  755         old_ldt_len = 0;
  756 
  757         /* Allocate a user ldt. */
  758         if ((pldt = mdp->md_ldt) == NULL || len > pldt->ldt_len) {
  759                 new_ldt = user_ldt_alloc(mdp, len);
  760                 if (new_ldt == NULL)
  761                         return (ENOMEM);
  762                 pldt = mdp->md_ldt;
  763 
  764                 if (pldt != NULL) {
  765                         if (new_ldt->ldt_len <= pldt->ldt_len) {
  766                                 /*
  767                                  * We just lost the race for allocation, so
  768                                  * free the new object and return.
  769                                  */
  770                                 mtx_unlock_spin(&dt_lock);
  771                                 pmap_trm_free(new_ldt->ldt_base,
  772                                    new_ldt->ldt_len * sizeof(union descriptor));
  773                                 free(new_ldt, M_SUBPROC);
  774                                 mtx_lock_spin(&dt_lock);
  775                                 return (0);
  776                         }
  777 
  778                         /*
  779                          * We have to substitute the current LDT entry for
  780                          * curproc with the new one since its size grew.
  781                          */
  782                         old_ldt_base = pldt->ldt_base;
  783                         old_ldt_len = pldt->ldt_len;
  784                         pldt->ldt_sd = new_ldt->ldt_sd;
  785                         pldt->ldt_base = new_ldt->ldt_base;
  786                         pldt->ldt_len = new_ldt->ldt_len;
  787                 } else
  788                         mdp->md_ldt = pldt = new_ldt;
  789 #ifdef SMP
  790                 /*
  791                  * Signal other cpus to reload ldt.  We need to unlock dt_lock
  792                  * here because other CPU will contest on it since their
  793                  * curthreads won't hold the lock and will block when trying
  794                  * to acquire it.
  795                  */
  796                 mtx_unlock_spin(&dt_lock);
  797                 smp_rendezvous(NULL, set_user_ldt_rv, NULL,
  798                     td->td_proc->p_vmspace);
  799 #else
  800                 set_user_ldt_locked(&td->td_proc->p_md);
  801                 mtx_unlock_spin(&dt_lock);
  802 #endif
  803                 if (old_ldt_base != NULL_LDT_BASE) {
  804                         pmap_trm_free(old_ldt_base, old_ldt_len *
  805                             sizeof(union descriptor));
  806                         free(new_ldt, M_SUBPROC);
  807                 }
  808                 mtx_lock_spin(&dt_lock);
  809         }
  810         return (0);
  811 }

Cache object: 24956f606524842d9d4a86ea5702b20a


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.