2 * Copyright (c) 2003,2004 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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
34 * $DragonFly: src/sys/kern/kern_xio.c,v 1.16 2008/05/09 07:24:45 dillon Exp $
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
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.
47 * TODO: check for busy page when modifying, check writeable.
50 #include <sys/param.h>
51 #include <sys/systm.h>
52 #include <sys/malloc.h>
54 #include <sys/vmmeter.h>
55 #include <sys/vnode.h>
58 #include <cpu/lwbuf.h>
61 #include <vm/vm_param.h>
63 #include <vm/vm_kern.h>
65 #include <vm/vm_map.h>
66 #include <vm/vm_object.h>
67 #include <vm/vm_page.h>
68 #include <vm/vm_pageout.h>
69 #include <vm/vm_pager.h>
70 #include <vm/vm_extern.h>
71 #include <vm/vm_page2.h>
74 * Just do basic initialization of an empty XIO
84 xio->xio_pages = xio->xio_internal_pages;
88 * Initialize an XIO given a userspace buffer. 0 is returned on success,
89 * an error code on failure. The actual number of bytes that could be
90 * accomodated in the XIO will be stored in xio_bytes and the page offset
91 * will be stored in xio_offset.
94 xio_init_ubuf(xio_t xio, void *ubase, size_t ubytes, int flags)
104 addr = trunc_page((vm_offset_t)ubase);
105 xio->xio_flags = flags;
112 vmprot = (flags & XIOF_WRITE) ? VM_PROT_WRITE : VM_PROT_READ;
113 xio->xio_offset = (vm_offset_t)ubase & PAGE_MASK;
114 xio->xio_pages = xio->xio_internal_pages;
115 if ((n = PAGE_SIZE - xio->xio_offset) > ubytes)
118 for (i = 0; n && i < XIO_INTERNAL_PAGES; ++i) {
119 m = vm_fault_page_quick(addr, vmprot, &error);
122 xio->xio_pages[i] = m;
125 if ((n = ubytes) > PAGE_SIZE)
130 * Check linearity, used by syslink to memory map DMA buffers.
132 if (flags & XIOF_VMLINEAR) {
136 if (m->object != m0->object || m->pindex != m0->pindex + i) {
145 * If a failure occured clean out what we loaded and return EFAULT.
146 * Return 0 on success. Do not dirty the pages.
148 if (i < XIO_INTERNAL_PAGES && n) {
149 xio->xio_flags &= ~XIOF_WRITE;
151 xio->xio_error = EFAULT;
154 return(xio->xio_error);
158 * Initialize an XIO given a kernelspace buffer. 0 is returned on success,
159 * an error code on failure. The actual number of bytes that could be
160 * accomodated in the XIO will be stored in xio_bytes and the page offset
161 * will be stored in xio_offset.
164 xio_init_kbuf(xio_t xio, void *kbase, size_t kbytes)
172 addr = trunc_page((vm_offset_t)kbase);
174 xio->xio_offset = (vm_offset_t)kbase & PAGE_MASK;
176 xio->xio_pages = xio->xio_internal_pages;
178 if ((n = PAGE_SIZE - xio->xio_offset) > kbytes)
180 lwkt_gettoken(&vm_token);
182 for (i = 0; n && i < XIO_INTERNAL_PAGES; ++i) {
183 if ((paddr = pmap_kextract(addr)) == 0)
185 m = PHYS_TO_VM_PAGE(paddr);
187 xio->xio_pages[i] = m;
190 if ((n = kbytes) > PAGE_SIZE)
195 lwkt_reltoken(&vm_token);
199 * If a failure occured clean out what we loaded and return EFAULT.
200 * Return 0 on success.
202 if (i < XIO_INTERNAL_PAGES && n) {
204 xio->xio_error = EFAULT;
206 return(xio->xio_error);
210 * Initialize an XIO given an array of vm_page pointers. The caller is
211 * responsible for any modified state changes for the pages.
214 xio_init_pages(xio_t xio, struct vm_page **mbase, int npages, int xflags)
218 KKASSERT(npages <= XIO_INTERNAL_PAGES);
220 xio->xio_flags = xflags;
222 xio->xio_bytes = npages * PAGE_SIZE;
223 xio->xio_pages = xio->xio_internal_pages;
224 xio->xio_npages = npages;
226 lwkt_gettoken(&vm_token);
228 for (i = 0; i < npages; ++i) {
229 vm_page_hold(mbase[i]);
230 xio->xio_pages[i] = mbase[i];
233 lwkt_reltoken(&vm_token);
238 * Cleanup an XIO so it can be destroyed. The pages associated with the
242 xio_release(xio_t xio)
247 lwkt_gettoken(&vm_token);
249 for (i = 0; i < xio->xio_npages; ++i) {
250 m = xio->xio_pages[i];
251 if (xio->xio_flags & XIOF_WRITE)
256 lwkt_reltoken(&vm_token);
260 xio->xio_error = ENOBUFS;
264 * Copy data between an XIO and a UIO. If the UIO represents userspace it
265 * must be relative to the current context.
267 * uoffset is the abstracted starting offset in the XIO, not the actual
268 * offset, and usually starts at 0.
270 * The XIO is not modified. The UIO is updated to reflect the copy.
272 * UIO_READ xio -> uio
273 * UIO_WRITE uio -> xio
276 xio_uio_copy(xio_t xio, int uoffset, struct uio *uio, size_t *sizep)
281 bytes = xio->xio_bytes - uoffset;
282 if (bytes > uio->uio_resid)
283 bytes = uio->uio_resid;
284 KKASSERT(bytes >= 0);
285 error = uiomove_fromphys(xio->xio_pages, xio->xio_offset + uoffset,
295 * Copy the specified number of bytes from the xio to a userland
296 * buffer. Return an error code or 0 on success.
298 * uoffset is the abstracted starting offset in the XIO, not the actual
299 * offset, and usually starts at 0.
301 * The XIO is not modified.
304 xio_copy_xtou(xio_t xio, int uoffset, void *uptr, int bytes)
312 struct lwbuf lwb_cache;
314 if (uoffset + bytes > xio->xio_bytes)
317 offset = (xio->xio_offset + uoffset) & PAGE_MASK;
318 if ((n = PAGE_SIZE - offset) > bytes)
322 for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT;
326 m = xio->xio_pages[i];
327 lwb = lwbuf_alloc(m, &lwb_cache);
328 error = copyout((char *)lwbuf_kva(lwb) + offset, uptr, n);
333 uptr = (char *)uptr + n;
336 if ((n = bytes) > PAGE_SIZE)
344 * Copy the specified number of bytes from the xio to a kernel
345 * buffer. Return an error code or 0 on success.
347 * uoffset is the abstracted starting offset in the XIO, not the actual
348 * offset, and usually starts at 0.
350 * The XIO is not modified.
353 xio_copy_xtok(xio_t xio, int uoffset, void *kptr, int bytes)
361 struct lwbuf lwb_cache;
363 if (bytes + uoffset > xio->xio_bytes)
366 offset = (xio->xio_offset + uoffset) & PAGE_MASK;
367 if ((n = PAGE_SIZE - offset) > bytes)
371 for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT;
375 m = xio->xio_pages[i];
376 lwb = lwbuf_alloc(m, &lwb_cache);
377 bcopy((char *)lwbuf_kva(lwb) + offset, kptr, n);
380 kptr = (char *)kptr + n;
383 if ((n = bytes) > PAGE_SIZE)
391 * Copy the specified number of bytes from userland to the xio.
392 * Return an error code or 0 on success.
394 * uoffset is the abstracted starting offset in the XIO, not the actual
395 * offset, and usually starts at 0.
397 * Data in pages backing the XIO will be modified.
400 xio_copy_utox(xio_t xio, int uoffset, const void *uptr, int bytes)
408 struct lwbuf lwb_cache;
410 if (uoffset + bytes > xio->xio_bytes)
413 offset = (xio->xio_offset + uoffset) & PAGE_MASK;
414 if ((n = PAGE_SIZE - offset) > bytes)
418 for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT;
422 m = xio->xio_pages[i];
423 lwb = lwbuf_alloc(m, &lwb_cache);
424 error = copyin(uptr, (char *)lwbuf_kva(lwb) + offset, n);
429 uptr = (const char *)uptr + n;
432 if ((n = bytes) > PAGE_SIZE)
440 * Copy the specified number of bytes from the kernel to the xio.
441 * Return an error code or 0 on success.
443 * uoffset is the abstracted starting offset in the XIO, not the actual
444 * offset, and usually starts at 0.
446 * Data in pages backing the XIO will be modified.
449 xio_copy_ktox(xio_t xio, int uoffset, const void *kptr, int bytes)
457 struct lwbuf lwb_cache;
459 if (uoffset + bytes > xio->xio_bytes)
462 offset = (xio->xio_offset + uoffset) & PAGE_MASK;
463 if ((n = PAGE_SIZE - offset) > bytes)
467 for (i = (xio->xio_offset + uoffset) >> PAGE_SHIFT;
471 m = xio->xio_pages[i];
472 lwb = lwbuf_alloc(m, &lwb_cache);
473 bcopy(kptr, (char *)lwbuf_kva(lwb) + offset, n);
476 kptr = (const char *)kptr + n;
479 if ((n = bytes) > PAGE_SIZE)