Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
NetBSD/sys/ufs/ext2fs/ext2fs_readwrite.c

346 lines
9.1 KiB
C
Raw Normal View History

a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
/* $NetBSD: ext2fs_readwrite.c,v 1.17 2001/09/15 20:36:41 chs Exp $ */
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
/*-
* Copyright (c) 1997 Manuel Bouyer.
* Copyright (c) 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)ufs_readwrite.c 8.8 (Berkeley) 8/4/94
* Modified for ext2fs by Manuel Bouyer.
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/resourcevar.h>
#include <sys/kernel.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/buf.h>
#include <sys/proc.h>
#include <sys/conf.h>
#include <sys/mount.h>
#include <sys/vnode.h>
#include <sys/malloc.h>
#include <sys/signalvar.h>
#include <ufs/ufs/quota.h>
#include <ufs/ufs/inode.h>
#include <ufs/ext2fs/ext2fs.h>
#include <ufs/ext2fs/ext2fs_extern.h>
#define doclusterread 0 /* XXX underway */
#define doclusterwrite 0
/*
* Vnode op for reading.
*/
/* ARGSUSED */
int
ext2fs_read(v)
void *v;
{
struct vop_read_args /* {
struct vnode *a_vp;
struct uio *a_uio;
int a_ioflag;
struct ucred *a_cred;
} */ *ap = v;
Remove register declarations.
2000-03-30 16:41:09 +04:00
struct vnode *vp;
struct inode *ip;
struct uio *uio;
struct m_ext2fs *fs;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
struct buf *bp;
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
void *win;
vsize_t bytelen;
Merge with Lite2 + local changes
1998-03-01 05:20:01 +03:00
ufs_daddr_t lbn, nextlbn;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
off_t bytesinfile;
long size, xfersize, blkoffset;
int error;
vp = ap->a_vp;
ip = VTOI(vp);
uio = ap->a_uio;
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_READ)
panic("%s: mode", "ext2fs_read");
if (vp->v_type == VLNK) {
if ((int)ip->i_e2fs_size < vp->v_mount->mnt_maxsymlinklen ||
(vp->v_mount->mnt_maxsymlinklen == 0 &&
ip->i_e2fs_nblock == 0))
panic("%s: short symlink", "ext2fs_read");
} else if (vp->v_type != VREG && vp->v_type != VDIR)
panic("%s: type %d", "ext2fs_read", vp->v_type);
#endif
fs = ip->i_e2fs;
if ((u_int64_t)uio->uio_offset >
((u_int64_t)0x80000000 * fs->e2fs_bsize - 1))
return (EFBIG);
if (uio->uio_resid == 0)
return (0);
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
if (vp->v_type == VREG) {
error = 0;
while (uio->uio_resid > 0) {
min() -> MIN(), max() -> MAX(). fixes more problems with file offsets > 4GB.
2001-02-27 07:37:44 +03:00
bytelen = MIN(ip->i_e2fs_size - uio->uio_offset,
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
uio->uio_resid);
if (bytelen == 0) {
break;
}
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
&bytelen, UBC_READ);
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
error = uiomove(win, bytelen, uio);
ubc_release(win, 0);
if (error) {
break;
}
}
goto out;
}
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
if ((bytesinfile = ip->i_e2fs_size - uio->uio_offset) <= 0)
break;
lbn = lblkno(fs, uio->uio_offset);
nextlbn = lbn + 1;
size = fs->e2fs_bsize;
blkoffset = blkoff(fs, uio->uio_offset);
xfersize = fs->e2fs_bsize - blkoffset;
if (uio->uio_resid < xfersize)
xfersize = uio->uio_resid;
if (bytesinfile < xfersize)
xfersize = bytesinfile;
if (lblktosize(fs, nextlbn) >= ip->i_e2fs_size)
error = bread(vp, lbn, size, NOCRED, &bp);
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
else {
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
int nextsize = fs->e2fs_bsize;
error = breadn(vp, lbn,
size, &nextlbn, &nextsize, 1, NOCRED, &bp);
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
}
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
if (error)
break;
/*
* We should only get non-zero b_resid when an I/O error
* has occurred, which should cause us to break above.
* However, if the short read did not cause an error,
* then we want to ensure that we do not uiomove bad
* or uninitialized data.
*/
size -= bp->b_resid;
if (size < xfersize) {
if (size == 0)
break;
xfersize = size;
}
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
if (error)
break;
brelse(bp);
}
if (bp != NULL)
brelse(bp);
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
out:
Implement support for IEEE Std 1003.1b-1993 synchronous I/O: * in the read vnode operator, check for IO_SYNC being set in the ioflag and synchronously update the file's meta-data if appropriate. * in the write vnode operator, update the appropriate checks for IO_SYNC being set in the ioflag to reflect that IO_DSYNC is now inclusive-or'ed into IO_SYNC, and require all IO_SYNC bits to be set for operations defined by synchronized I/O file integrity completion but not by synchronized I/O data integrity completion.
1998-08-02 22:57:23 +04:00
if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) {
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
ip->i_flag |= IN_ACCESS;
Pass null pointers to VOP_UPDATE rather than having all the callers fetch the current time themselves.
1999-03-06 00:09:48 +03:00
if ((ap->a_ioflag & IO_SYNC) == IO_SYNC)
Change the sementics of the last parameter from a boolean ("waitfor") to a set of flags ("flags"). Two flags are defined, UPDATE_WAIT and UPDATE_DIROP. Under the old semantics, VOP_UPDATE would block if waitfor were set, under the assumption that directory operations should be done synchronously. At least LFS and FFS+softdep do not make this assumption; FFS+softdep got around the problem by enclosing all relevant calls to VOP_UPDATE in a "if(!DOINGSOFTDEP(vp))", while LFS simply ignored waitfor, one of the reasons why NFS-serving an LFS filesystem did not work properly. Under the new semantics, the UPDATE_DIROP flag is a hint to the fs-specific update routine that the call comes from a dirop routine, and should be wait for, or not, accordingly. Closes PR#8996.
2000-05-14 03:43:06 +04:00
error = VOP_UPDATE(vp, NULL, NULL, UPDATE_WAIT);
Implement support for IEEE Std 1003.1b-1993 synchronous I/O: * in the read vnode operator, check for IO_SYNC being set in the ioflag and synchronously update the file's meta-data if appropriate. * in the write vnode operator, update the appropriate checks for IO_SYNC being set in the ioflag to reflect that IO_DSYNC is now inclusive-or'ed into IO_SYNC, and require all IO_SYNC bits to be set for operations defined by synchronized I/O file integrity completion but not by synchronized I/O data integrity completion.
1998-08-02 22:57:23 +04:00
}
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
return (error);
}
/*
* Vnode op for writing.
*/
int
ext2fs_write(v)
void *v;
{
struct vop_write_args /* {
struct vnode *a_vp;
struct uio *a_uio;
int a_ioflag;
struct ucred *a_cred;
} */ *ap = v;
Remove register declarations.
2000-03-30 16:41:09 +04:00
struct vnode *vp;
struct uio *uio;
struct inode *ip;
struct m_ext2fs *fs;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
struct buf *bp;
struct proc *p;
Merge with Lite2 + local changes
1998-03-01 05:20:01 +03:00
ufs_daddr_t lbn;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
off_t osize;
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
int blkoffset, error, flags, ioflag, resid, xfersize;
vsize_t bytelen;
void *win;
off_t oldoff;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
ioflag = ap->a_ioflag;
uio = ap->a_uio;
vp = ap->a_vp;
ip = VTOI(vp);
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
error = 0;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
#ifdef DIAGNOSTIC
if (uio->uio_rw != UIO_WRITE)
panic("%s: mode", "ext2fs_write");
#endif
switch (vp->v_type) {
case VREG:
if (ioflag & IO_APPEND)
uio->uio_offset = ip->i_e2fs_size;
if ((ip->i_e2fs_flags & EXT2_APPEND) &&
uio->uio_offset != ip->i_e2fs_size)
return (EPERM);
/* FALLTHROUGH */
case VLNK:
break;
case VDIR:
if ((ioflag & IO_SYNC) == 0)
panic("%s: nonsync dir write", "ext2fs_write");
break;
default:
panic("%s: type", "ext2fs_write");
}
fs = ip->i_e2fs;
if (uio->uio_offset < 0 ||
(u_int64_t)uio->uio_offset + uio->uio_resid >
((u_int64_t)0x80000000 * fs->e2fs_bsize - 1))
return (EFBIG);
/*
* Maybe this should be above the vnode op call, but so long as
* file servers have no limits, I don't think it matters.
*/
p = uio->uio_procp;
if (vp->v_type == VREG && p &&
uio->uio_offset + uio->uio_resid >
p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
psignal(p, SIGXFSZ);
return (EFBIG);
}
resid = uio->uio_resid;
osize = ip->i_e2fs_size;
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
if (vp->v_type == VREG) {
while (uio->uio_resid > 0) {
oldoff = uio->uio_offset;
blkoffset = blkoff(fs, uio->uio_offset);
min() -> MIN(), max() -> MAX(). fixes more problems with file offsets > 4GB.
2001-02-27 07:37:44 +03:00
bytelen = MIN(fs->e2fs_bsize - blkoffset,
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
uio->uio_resid);
error = ext2fs_balloc_range(vp, uio->uio_offset,
bytelen, ap->a_cred, 0);
if (error) {
break;
}
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
win = ubc_alloc(&vp->v_uobj, uio->uio_offset,
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
&bytelen, UBC_WRITE);
error = uiomove(win, bytelen, uio);
ubc_release(win, 0);
if (error) {
break;
}
/*
* flush what we just wrote if necessary.
* XXXUBC simplistic async flushing.
*/
if (oldoff >> 16 != uio->uio_offset >> 16) {
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
simple_lock(&vp->v_uobj.vmobjlock);
error = vp->v_uobj.pgops->pgo_put(
&vp->v_uobj, (oldoff >> 16) << 16,
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
(uio->uio_offset >> 16) << 16, PGO_CLEANIT);
}
}
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
if (error == 0 && ioflag & IO_SYNC) {
simple_lock(&vp->v_uobj.vmobjlock);
error = vp->v_uobj.pgops->pgo_put(&vp->v_uobj, oldoff,
oldoff + bytelen, PGO_CLEANIT|PGO_SYNCIO);
}
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
goto out;
}
flags = ioflag & IO_SYNC ? B_SYNC : 0;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
for (error = 0; uio->uio_resid > 0;) {
lbn = lblkno(fs, uio->uio_offset);
blkoffset = blkoff(fs, uio->uio_offset);
min() -> MIN(), max() -> MAX(). fixes more problems with file offsets > 4GB.
2001-02-27 07:37:44 +03:00
xfersize = MIN(fs->e2fs_bsize - blkoffset, uio->uio_resid);
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
if (xfersize < fs->e2fs_bsize)
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
fix merge error: ext2fs uses a custom balloc rather than a VOP-style one.
2000-12-01 10:02:40 +03:00
error = ext2fs_balloc(ip,
lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
if (error)
break;
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
if (ip->i_e2fs_size < uio->uio_offset + xfersize) {
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
ip->i_e2fs_size = uio->uio_offset + xfersize;
}
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio);
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
if (ioflag & IO_SYNC)
(void)bwrite(bp);
else if (xfersize + blkoffset == fs->e2fs_bsize)
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
bawrite(bp);
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
else
bdwrite(bp);
if (error || xfersize == 0)
break;
}
/*
* If we successfully wrote any data, and we are not the superuser
* we clear the setuid and setgid bits as a precaution against
* tampering.
*/
Initial integration of the Unified Buffer Cache project.
2000-11-27 11:39:39 +03:00
out:
ip->i_flag |= IN_CHANGE | IN_UPDATE;
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
if (resid > uio->uio_resid && ap->a_cred && ap->a_cred->cr_uid != 0)
ip->i_e2fs_mode &= ~(ISUID | ISGID);
if (error) {
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
(void) VOP_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred,
uio->uio_procp);
uio->uio_offset -= resid - uio->uio_resid;
uio->uio_resid = resid;
Pass null pointers to VOP_UPDATE rather than having all the callers fetch the current time themselves.
1999-03-06 00:09:48 +03:00
} else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC)
Change the sementics of the last parameter from a boolean ("waitfor") to a set of flags ("flags"). Two flags are defined, UPDATE_WAIT and UPDATE_DIROP. Under the old semantics, VOP_UPDATE would block if waitfor were set, under the assumption that directory operations should be done synchronously. At least LFS and FFS+softdep do not make this assumption; FFS+softdep got around the problem by enclosing all relevant calls to VOP_UPDATE in a "if(!DOINGSOFTDEP(vp))", while LFS simply ignored waitfor, one of the reasons why NFS-serving an LFS filesystem did not work properly. Under the new semantics, the UPDATE_DIROP flag is a hint to the fs-specific update routine that the call comes from a dirop routine, and should be wait for, or not, accordingly. Closes PR#8996.
2000-05-14 03:43:06 +04:00
error = VOP_UPDATE(vp, NULL, NULL, UPDATE_WAIT);
a whole bunch of changes to improve performance and robustness under load: - remove special treatment of pager_map mappings in pmaps. this is required now, since I've removed the globals that expose the address range. pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's no longer any need to special-case it. - eliminate struct uvm_vnode by moving its fields into struct vnode. - rewrite the pageout path. the pager is now responsible for handling the high-level requests instead of only getting control after a bunch of work has already been done on its behalf. this will allow us to UBCify LFS, which needs tighter control over its pages than other filesystems do. writing a page to disk no longer requires making it read-only, which allows us to write wired pages without causing all kinds of havoc. - use a new PG_PAGEOUT flag to indicate that a page should be freed on behalf of the pagedaemon when it's unlocked. this flag is very similar to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the pageout fails due to eg. an indirect-block buffer being locked. this allows us to remove the "version" field from struct vm_page, and together with shrinking "loan_count" from 32 bits to 16, struct vm_page is now 4 bytes smaller. - no longer use PG_RELEASED for swap-backed pages. if the page is busy because it's being paged out, we can't release the swap slot to be reallocated until that write is complete, but unlike with vnodes we don't keep a count of in-progress writes so there's no good way to know when the write is done. instead, when we need to free a busy swap-backed page, just sleep until we can get it busy ourselves. - implement a fast-path for extending writes which allows us to avoid zeroing new pages. this substantially reduces cpu usage. - encapsulate the data used by the genfs code in a struct genfs_node, which must be the first element of the filesystem-specific vnode data for filesystems which use genfs_{get,put}pages(). - eliminate many of the UVM pagerops, since they aren't needed anymore now that the pager "put" operation is a higher-level operation. - enhance the genfs code to allow NFS to use the genfs_{get,put}pages instead of a modified copy. - clean up struct vnode by removing all the fields that used to be used by the vfs_cluster.c code (which we don't use anymore with UBC). - remove kmem_object and mb_object since they were useless. instead of allocating pages to these objects, we now just allocate pages with no object. such pages are mapped in the kernel until they are freed, so we can use the mapping to find the page to free it. this allows us to remove splvm() protection in several places. The sum of all these changes improves write throughput on my decstation 5000/200 to within 1% of the rate of NetBSD 1.5 and reduces the elapsed time for "make release" of a NetBSD 1.5 source tree on my 128MB pc to 10% less than a 1.5 kernel took.
2001-09-16 00:36:31 +04:00
KASSERT(vp->v_size == ip->i_e2fs_size);
The ext2fs layer, based on the ffs/ufs one. Uses a few functions from sys/ufs/ufs/
1997-06-11 13:33:37 +04:00
return (error);
}