sys/kern/kern_xio.c

   1 /*
   2  * Copyright (c) 2003,2004 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   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  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * $DragonFly: src/sys/kern/kern_xio.c,v 1.10 2005/06/06 15:02:28 dillon Exp $
  35  */
  36 /*
  37  * Kernel XIO interface.  An initialized XIO is basically a collection of
  38  * appropriately held vm_page_t's.  XIO buffers are vmspace agnostic and
  39  * can represent userspace or kernelspace buffers, and can be passed to
  40  * foreign threads outside of the originating vmspace.  XIO buffers are
  41  * not mapped into KVM and thus can be manipulated and passed around with
  42  * very low overheads.
  43  *
  44  * The intent is for XIO to be used in the I/O path, VFS, CAPS, and other
  45  * places that need to pass (possibly userspace) data between threads.
  46  *
  47  * TODO: check for busy page when modifying, check writeable.
  48  */
  49
  50 #include <sys/param.h>
  51 #include <sys/systm.h>
  52 #include <sys/malloc.h>
  53 #include <sys/proc.h>
  54 #include <sys/vmmeter.h>
  55 #include <sys/vnode.h>
  56 #include <sys/xio.h>
  57 #include <sys/sfbuf.h>
  58
  59 #include <vm/vm.h>
  60 #include <vm/vm_param.h>
  61 #include <sys/lock.h>
  62 #include <vm/vm_kern.h>
  63 #include <vm/pmap.h>
  64 #include <vm/vm_map.h>
  65 #include <vm/vm_object.h>
  66 #include <vm/vm_page.h>
  67 #include <vm/vm_pageout.h>
  68 #include <vm/vm_pager.h>
  69 #include <vm/vm_extern.h>
  70 #include <vm/vm_page2.h>
  71
  72 /*
  73  * Just do basic initialization of an empty XIO
  74  */
  75 void
  76 xio_init(xio_t xio)
  77 {
  78     xio->xio_flags = 0;
  79     xio->xio_bytes = 0;
  80     xio->xio_error = 0;
  81     xio->xio_offset = 0;
  82     xio->xio_npages = 0;
  83     xio->xio_pages = xio->xio_internal_pages;
  84 }
  85
  86 /*
  87  * Initialize an XIO given a userspace buffer.  0 is returned on success,
  88  * an error code on failure.  The actual number of bytes that could be
  89  * accomodated in the XIO will be stored in xio_bytes and the page offset
  90  * will be stored in xio_offset.
  91  */
  92 int
  93 xio_init_ubuf(xio_t xio, void *ubase, size_t ubytes, int flags)
  94 {
  95     vm_offset_t addr;
  96     vm_paddr_t paddr;
  97     vm_page_t m;
  98     int i;
  99     int n;
 100     int vmprot;
 101
 102     addr = trunc_page((vm_offset_t)ubase);
 103     xio->xio_flags = flags;
 104     xio->xio_bytes = 0;
 105     xio->xio_error = 0;
 106     if (ubytes == 0) {
 107         xio->xio_offset = 0;
 108         xio->xio_npages = 0;
 109     } else {
 110         vmprot = (flags & XIOF_WRITE) ? VM_PROT_WRITE : VM_PROT_READ;
 111         xio->xio_offset = (vm_offset_t)ubase & PAGE_MASK;
 112         xio->xio_pages = xio->xio_internal_pages;
 113         if ((n = PAGE_SIZE - xio->xio_offset) > ubytes)
 114             n = ubytes;
 115         for (i = 0; n && i < XIO_INTERNAL_PAGES; ++i) {
 116             if (vm_fault_quick((caddr_t)addr, vmprot) < 0)
 117                 break;
 118             if ((paddr = pmap_kextract(addr)) == 0)
 119                 break;
 120             crit_enter();
 121             m = PHYS_TO_VM_PAGE(paddr);
 122             vm_page_hold(m);
 123             crit_exit();
 124             xio->xio_pages[i] = m;
 125             ubytes -= n;
 126             xio->xio_bytes += n;
 127             if ((n = ubytes) > PAGE_SIZE)
 128                 n = PAGE_SIZE;
 129             addr += PAGE_SIZE;
 130         }
 131         xio->xio_npages = i;
 132
 133         /*
 134          * If a failure occured clean out what we loaded and return EFAULT.
 135          * Return 0 on success.
 136          */
 137         if (i < XIO_INTERNAL_PAGES && n) {
 138             xio_release(xio);
 139             xio->xio_error = EFAULT;
 140         }
 141     }
 142     return(xio->xio_error);
 143 }
 144
 145 /*
 146  * Initialize an XIO given a kernelspace buffer.  0 is returned on success,
 147  * an error code on failure.  The actual number of bytes that could be
 148  * accomodated in the XIO will be stored in xio_bytes and the page offset
 149  * will be stored in xio_offset.
 150  */
 151 int
 152 xio_init_kbuf(xio_t xio, void *kbase, size_t kbytes)
 153 {
 154     vm_offset_t addr;
 155     vm_paddr_t paddr;
 156     vm_page_t m;
 157     int i;
 158     int n;
 159
 160     addr = trunc_page((vm_offset_t)kbase);
 161     xio->xio_flags = 0;
 162     xio->xio_offset = (vm_offset_t)kbase & PAGE_MASK;
 163     xio->xio_bytes = 0;
 164     xio->xio_pages = xio->xio_internal_pages;
 165     xio->xio_error = 0;
 166     if ((n = PAGE_SIZE - xio->xio_offset) > kbytes)
 167         n = kbytes;
 168     for (i = 0; n && i < XIO_INTERNAL_PAGES; ++i) {
 169         if ((paddr = pmap_kextract(addr)) == 0)
 170             break;
 171         crit_enter();
 172         m = PHYS_TO_VM_PAGE(paddr);
 173         vm_page_hold(m);
 174         crit_exit();
 175         xio->xio_pages[i] = m;
 176         kbytes -= n;
 177         xio->xio_bytes += n;
 178         if ((n = kbytes) > PAGE_SIZE)
 179             n = PAGE_SIZE;
 180         addr += PAGE_SIZE;
 181     }
 182     xio->xio_npages = i;
 183
 184     /*
 185      * If a failure occured clean out what we loaded and return EFAULT.
 186      * Return 0 on success.
 187      */
 188     if (i < XIO_INTERNAL_PAGES && n) {
 189         xio_release(xio);
 190         xio->xio_error = EFAULT;
 191     }
 192     return(xio->xio_error);
 193 }
 194
 195 /*
 196  * Cleanup an XIO so it can be destroyed.  The pages associated with the
 197  * XIO are released.
 198  */
 199 void
 200 xio_release(xio_t xio)
 201 {
 202     int i;
 203     vm_page_t m;
 204
 205     crit_enter();
 206     for (i = 0; i < xio->xio_npages; ++i) {
 207         m = xio->xio_pages[i];
 208         vm_page_unhold(m);
 209     }
 210     crit_exit();
 211     xio->xio_offset = 0;
 212     xio->xio_npages = 0;
 213     xio->xio_bytes = 0;
 214     xio->xio_error = ENOBUFS;
 215 }
 216
 217 /*
 218  * Copy data between an XIO and a UIO.  If the UIO represents userspace it
 219  * must be relative to the current context.
 220  *
 221  * uoffset is the abstracted starting offset in the XIO, not the actual
 222  * offset, and usually starts at 0.
 223  *
 224  * The XIO is not modified.  The UIO is updated to reflect the copy.
 225  *
 226  * UIO_READ     xio -> uio
 227  * UIO_WRITE    uio -> xio
 228  */
 229 int
 230 xio_uio_copy(xio_t xio, int uoffset, struct uio *uio, int *sizep)
 231 {
 232     int error;
 233     int bytes;
 234
 235     bytes = xio->xio_bytes - uoffset;
 236     if (bytes > uio->uio_resid)
 237         bytes = uio->uio_resid;
 238     KKASSERT(bytes >= 0);
 239     error = uiomove_fromphys(xio->xio_pages, xio->xio_offset + uoffset,
 240                                 bytes, uio);
 241     if (error == 0)
 242         *sizep = bytes;
 243     else
 244         *sizep = 0;
 245     return(error);
 246 }
 247
 248 /*
 249  * Copy the specified number of bytes from the xio to a userland
 250  * buffer.  Return an error code or 0 on success.
 251  *
 252  * uoffset is the abstracted starting offset in the XIO, not the actual
 253  * offset, and usually starts at 0.
 254  *
 255  * The XIO is not modified.
 256  */
 257 int
 258 xio_copy_xtou(xio_t xio, int uoffset, void *uptr, int bytes)
 259 {
 260     int i;
 261     int n;
 262     int error;
 263     int offset;
 264     vm_page_t m;
 265     struct sf_buf *sf;
 266
 267     if (bytes > xio->xio_bytes)
 268         return(EFAULT);
 269
 270     offset = (xio->xio_offset + uoffset) & PAGE_MASK;
 271     if ((n = PAGE_SIZE - offset) > bytes)
 272         n = bytes;
 273
 274     error = 0;
 275     for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT;
 276          i < xio->xio_npages;
 277          ++i
 278     ) {
 279         m = xio->xio_pages[i];
 280         sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
 281         error = copyout((char *)sf_buf_kva(sf) + offset, uptr, n);
 282         sf_buf_free(sf);
 283         if (error)
 284             break;
 285         bytes -= n;
 286         uptr = (char *)uptr + n;
 287         if (bytes == 0)
 288             break;
 289         if ((n = bytes) > PAGE_SIZE)
 290             n = PAGE_SIZE;
 291         offset = 0;
 292     }
 293     return(error);
 294 }
 295
 296 /*
 297  * Copy the specified number of bytes from the xio to a kernel
 298  * buffer.  Return an error code or 0 on success.
 299  *
 300  * uoffset is the abstracted starting offset in the XIO, not the actual
 301  * offset, and usually starts at 0.
 302  *
 303  * The XIO is not modified.
 304  */
 305 int
 306 xio_copy_xtok(xio_t xio, int uoffset, void *kptr, int bytes)
 307 {
 308     int i;
 309     int n;
 310     int error;
 311     int offset;
 312     vm_page_t m;
 313     struct sf_buf *sf;
 314
 315     if (bytes + uoffset > xio->xio_bytes)
 316         return(EFAULT);
 317
 318     offset = (xio->xio_offset + uoffset) & PAGE_MASK;
 319     if ((n = PAGE_SIZE - offset) > bytes)
 320         n = bytes;
 321
 322     error = 0;
 323     for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT;
 324          i < xio->xio_npages;
 325          ++i
 326     ) {
 327         m = xio->xio_pages[i];
 328         sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
 329         bcopy((char *)sf_buf_kva(sf) + offset, kptr, n);
 330         sf_buf_free(sf);
 331         bytes -= n;
 332         kptr = (char *)kptr + n;
 333         if (bytes == 0)
 334             break;
 335         if ((n = bytes) > PAGE_SIZE)
 336             n = PAGE_SIZE;
 337         offset = 0;
 338     }
 339     return(error);
 340 }
 341