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 1.12 2007/01/11 20:53:41 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_page_t m; 97 int error; 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 m = vm_fault_page_quick(addr, vmprot, &error); 117 if (m == NULL) 118 break; 119 xio->xio_pages[i] = m; 120 ubytes -= n; 121 xio->xio_bytes += n; 122 if ((n = ubytes) > PAGE_SIZE) 123 n = PAGE_SIZE; 124 addr += PAGE_SIZE; 125 } 126 xio->xio_npages = i; 127 128 /* 129 * If a failure occured clean out what we loaded and return EFAULT. 130 * Return 0 on success. 131 */ 132 if (i < XIO_INTERNAL_PAGES && n) { 133 xio_release(xio); 134 xio->xio_error = EFAULT; 135 } 136 } 137 return(xio->xio_error); 138 } 139 140 /* 141 * Initialize an XIO given a kernelspace buffer. 0 is returned on success, 142 * an error code on failure. The actual number of bytes that could be 143 * accomodated in the XIO will be stored in xio_bytes and the page offset 144 * will be stored in xio_offset. 145 */ 146 int 147 xio_init_kbuf(xio_t xio, void *kbase, size_t kbytes) 148 { 149 vm_offset_t addr; 150 vm_paddr_t paddr; 151 vm_page_t m; 152 int i; 153 int n; 154 155 addr = trunc_page((vm_offset_t)kbase); 156 xio->xio_flags = 0; 157 xio->xio_offset = (vm_offset_t)kbase & PAGE_MASK; 158 xio->xio_bytes = 0; 159 xio->xio_pages = xio->xio_internal_pages; 160 xio->xio_error = 0; 161 if ((n = PAGE_SIZE - xio->xio_offset) > kbytes) 162 n = kbytes; 163 for (i = 0; n && i < XIO_INTERNAL_PAGES; ++i) { 164 if ((paddr = pmap_kextract(addr)) == 0) 165 break; 166 crit_enter(); 167 m = PHYS_TO_VM_PAGE(paddr); 168 vm_page_hold(m); 169 crit_exit(); 170 xio->xio_pages[i] = m; 171 kbytes -= n; 172 xio->xio_bytes += n; 173 if ((n = kbytes) > PAGE_SIZE) 174 n = PAGE_SIZE; 175 addr += PAGE_SIZE; 176 } 177 xio->xio_npages = i; 178 179 /* 180 * If a failure occured clean out what we loaded and return EFAULT. 181 * Return 0 on success. 182 */ 183 if (i < XIO_INTERNAL_PAGES && n) { 184 xio_release(xio); 185 xio->xio_error = EFAULT; 186 } 187 return(xio->xio_error); 188 } 189 190 /* 191 * Cleanup an XIO so it can be destroyed. The pages associated with the 192 * XIO are released. 193 */ 194 void 195 xio_release(xio_t xio) 196 { 197 int i; 198 vm_page_t m; 199 200 crit_enter(); 201 for (i = 0; i < xio->xio_npages; ++i) { 202 m = xio->xio_pages[i]; 203 vm_page_unhold(m); 204 } 205 crit_exit(); 206 xio->xio_offset = 0; 207 xio->xio_npages = 0; 208 xio->xio_bytes = 0; 209 xio->xio_error = ENOBUFS; 210 } 211 212 /* 213 * Copy data between an XIO and a UIO. If the UIO represents userspace it 214 * must be relative to the current context. 215 * 216 * uoffset is the abstracted starting offset in the XIO, not the actual 217 * offset, and usually starts at 0. 218 * 219 * The XIO is not modified. The UIO is updated to reflect the copy. 220 * 221 * UIO_READ xio -> uio 222 * UIO_WRITE uio -> xio 223 */ 224 int 225 xio_uio_copy(xio_t xio, int uoffset, struct uio *uio, int *sizep) 226 { 227 int error; 228 int bytes; 229 230 bytes = xio->xio_bytes - uoffset; 231 if (bytes > uio->uio_resid) 232 bytes = uio->uio_resid; 233 KKASSERT(bytes >= 0); 234 error = uiomove_fromphys(xio->xio_pages, xio->xio_offset + uoffset, 235 bytes, uio); 236 if (error == 0) 237 *sizep = bytes; 238 else 239 *sizep = 0; 240 return(error); 241 } 242 243 /* 244 * Copy the specified number of bytes from the xio to a userland 245 * buffer. Return an error code or 0 on success. 246 * 247 * uoffset is the abstracted starting offset in the XIO, not the actual 248 * offset, and usually starts at 0. 249 * 250 * The XIO is not modified. 251 */ 252 int 253 xio_copy_xtou(xio_t xio, int uoffset, void *uptr, int bytes) 254 { 255 int i; 256 int n; 257 int error; 258 int offset; 259 vm_page_t m; 260 struct sf_buf *sf; 261 262 if (bytes > xio->xio_bytes) 263 return(EFAULT); 264 265 offset = (xio->xio_offset + uoffset) & PAGE_MASK; 266 if ((n = PAGE_SIZE - offset) > bytes) 267 n = bytes; 268 269 error = 0; 270 for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT; 271 i < xio->xio_npages; 272 ++i 273 ) { 274 m = xio->xio_pages[i]; 275 sf = sf_buf_alloc(m, SFB_CPUPRIVATE); 276 error = copyout((char *)sf_buf_kva(sf) + offset, uptr, n); 277 sf_buf_free(sf); 278 if (error) 279 break; 280 bytes -= n; 281 uptr = (char *)uptr + n; 282 if (bytes == 0) 283 break; 284 if ((n = bytes) > PAGE_SIZE) 285 n = PAGE_SIZE; 286 offset = 0; 287 } 288 return(error); 289 } 290 291 /* 292 * Copy the specified number of bytes from the xio to a kernel 293 * buffer. Return an error code or 0 on success. 294 * 295 * uoffset is the abstracted starting offset in the XIO, not the actual 296 * offset, and usually starts at 0. 297 * 298 * The XIO is not modified. 299 */ 300 int 301 xio_copy_xtok(xio_t xio, int uoffset, void *kptr, int bytes) 302 { 303 int i; 304 int n; 305 int error; 306 int offset; 307 vm_page_t m; 308 struct sf_buf *sf; 309 310 if (bytes + uoffset > xio->xio_bytes) 311 return(EFAULT); 312 313 offset = (xio->xio_offset + uoffset) & PAGE_MASK; 314 if ((n = PAGE_SIZE - offset) > bytes) 315 n = bytes; 316 317 error = 0; 318 for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT; 319 i < xio->xio_npages; 320 ++i 321 ) { 322 m = xio->xio_pages[i]; 323 sf = sf_buf_alloc(m, SFB_CPUPRIVATE); 324 bcopy((char *)sf_buf_kva(sf) + offset, kptr, n); 325 sf_buf_free(sf); 326 bytes -= n; 327 kptr = (char *)kptr + n; 328 if (bytes == 0) 329 break; 330 if ((n = bytes) > PAGE_SIZE) 331 n = PAGE_SIZE; 332 offset = 0; 333 } 334 return(error); 335 } 336 337