Blame view
fs/xfs/xfs_filestream.c
8.68 KB
0b61f8a40
|
1 |
// SPDX-License-Identifier: GPL-2.0 |
2a82b8be8
|
2 3 |
/* * Copyright (c) 2006-2007 Silicon Graphics, Inc. |
2cd2ef6a3
|
4 |
* Copyright (c) 2014 Christoph Hellwig. |
2a82b8be8
|
5 |
* All Rights Reserved. |
2a82b8be8
|
6 7 |
*/ #include "xfs.h" |
5467b34bd
|
8 |
#include "xfs_shared.h" |
a4fbe6ab1
|
9 |
#include "xfs_format.h" |
239880ef6
|
10 11 |
#include "xfs_log_format.h" #include "xfs_trans_resv.h" |
239880ef6
|
12 13 |
#include "xfs_sb.h" #include "xfs_mount.h" |
2a82b8be8
|
14 |
#include "xfs_inode.h" |
2a82b8be8
|
15 16 |
#include "xfs_bmap.h" #include "xfs_alloc.h" |
2a82b8be8
|
17 |
#include "xfs_mru_cache.h" |
0b1b213fc
|
18 |
#include "xfs_trace.h" |
3fd129b63
|
19 |
#include "xfs_ag_resv.h" |
3e3673e30
|
20 |
#include "xfs_trans.h" |
f368b29ba
|
21 |
#include "xfs_filestream.h" |
2a82b8be8
|
22 |
|
2cd2ef6a3
|
23 24 |
struct xfs_fstrm_item { struct xfs_mru_cache_elem mru; |
2cd2ef6a3
|
25 26 27 28 29 30 31 |
xfs_agnumber_t ag; /* AG in use for this directory */ }; enum xfs_fstrm_alloc { XFS_PICK_USERDATA = 1, XFS_PICK_LOWSPACE = 2, }; |
2a82b8be8
|
32 |
|
0664ce8d0
|
33 34 |
/* * Allocation group filestream associations are tracked with per-ag atomic |
2cd2ef6a3
|
35 |
* counters. These counters allow xfs_filestream_pick_ag() to tell whether a |
b38e07401
|
36 |
* particular AG already has active filestreams associated with it. |
0664ce8d0
|
37 |
*/ |
b94acd478
|
38 |
int |
0664ce8d0
|
39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
xfs_filestream_peek_ag( xfs_mount_t *mp, xfs_agnumber_t agno) { struct xfs_perag *pag; int ret; pag = xfs_perag_get(mp, agno); ret = atomic_read(&pag->pagf_fstrms); xfs_perag_put(pag); return ret; } static int xfs_filestream_get_ag( xfs_mount_t *mp, xfs_agnumber_t agno) { struct xfs_perag *pag; int ret; pag = xfs_perag_get(mp, agno); ret = atomic_inc_return(&pag->pagf_fstrms); xfs_perag_put(pag); return ret; } static void xfs_filestream_put_ag( xfs_mount_t *mp, xfs_agnumber_t agno) { struct xfs_perag *pag; pag = xfs_perag_get(mp, agno); atomic_dec(&pag->pagf_fstrms); xfs_perag_put(pag); } |
2a82b8be8
|
77 |
|
2cd2ef6a3
|
78 79 |
static void xfs_fstrm_free_func( |
7fcd3efa1
|
80 |
void *data, |
2cd2ef6a3
|
81 82 |
struct xfs_mru_cache_elem *mru) { |
7fcd3efa1
|
83 |
struct xfs_mount *mp = data; |
2cd2ef6a3
|
84 85 |
struct xfs_fstrm_item *item = container_of(mru, struct xfs_fstrm_item, mru); |
7fcd3efa1
|
86 87 |
xfs_filestream_put_ag(mp, item->ag); trace_xfs_filestream_free(mp, mru->key, item->ag); |
2cd2ef6a3
|
88 |
|
1919adda0
|
89 |
kmem_free(item); |
2cd2ef6a3
|
90 |
} |
2a82b8be8
|
91 92 93 94 95 |
/* * Scan the AGs starting at startag looking for an AG that isn't in use and has * at least minlen blocks free. */ static int |
2cd2ef6a3
|
96 97 98 99 100 101 |
xfs_filestream_pick_ag( struct xfs_inode *ip, xfs_agnumber_t startag, xfs_agnumber_t *agp, int flags, xfs_extlen_t minlen) |
2a82b8be8
|
102 |
{ |
2cd2ef6a3
|
103 104 105 |
struct xfs_mount *mp = ip->i_mount; struct xfs_fstrm_item *item; struct xfs_perag *pag; |
b94acd478
|
106 |
xfs_extlen_t longest, free = 0, minfree, maxfree = 0; |
2cd2ef6a3
|
107 |
xfs_agnumber_t ag, max_ag = NULLAGNUMBER; |
2cd2ef6a3
|
108 |
int err, trylock, nscan; |
c19b3b05a
|
109 |
ASSERT(S_ISDIR(VFS_I(ip)->i_mode)); |
2a82b8be8
|
110 111 112 113 114 115 116 117 118 119 120 |
/* 2% of an AG's blocks must be free for it to be chosen. */ minfree = mp->m_sb.sb_agblocks / 50; ag = startag; *agp = NULLAGNUMBER; /* For the first pass, don't sleep trying to init the per-AG. */ trylock = XFS_ALLOC_FLAG_TRYLOCK; for (nscan = 0; 1; nscan++) { |
7fcd3efa1
|
121 |
trace_xfs_filestream_scan(mp, ip->i_ino, ag); |
b94acd478
|
122 |
|
4196ac08c
|
123 |
pag = xfs_perag_get(mp, ag); |
2a82b8be8
|
124 125 126 |
if (!pag->pagf_init) { err = xfs_alloc_pagf_init(mp, NULL, ag, trylock); |
f48e2df8a
|
127 |
if (err) { |
4196ac08c
|
128 |
xfs_perag_put(pag); |
f48e2df8a
|
129 130 131 132 |
if (err != -EAGAIN) return err; /* Couldn't lock the AGF, skip this AG. */ continue; |
4196ac08c
|
133 |
} |
2a82b8be8
|
134 |
} |
2a82b8be8
|
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 |
/* Keep track of the AG with the most free blocks. */ if (pag->pagf_freeblks > maxfree) { maxfree = pag->pagf_freeblks; max_ag = ag; } /* * The AG reference count does two things: it enforces mutual * exclusion when examining the suitability of an AG in this * loop, and it guards against two filestreams being established * in the same AG as each other. */ if (xfs_filestream_get_ag(mp, ag) > 1) { xfs_filestream_put_ag(mp, ag); goto next_ag; } |
a1f69417c
|
151 |
longest = xfs_alloc_longest_free_extent(pag, |
3fd129b63
|
152 153 |
xfs_alloc_min_freelist(mp, pag), xfs_ag_resv_needed(pag, XFS_AG_RESV_NONE)); |
2a82b8be8
|
154 155 156 157 158 159 160 |
if (((minlen && longest >= minlen) || (!minlen && pag->pagf_freeblks >= minfree)) && (!pag->pagf_metadata || !(flags & XFS_PICK_USERDATA) || (flags & XFS_PICK_LOWSPACE))) { /* Break out, retaining the reference on the AG. */ free = pag->pagf_freeblks; |
4196ac08c
|
161 |
xfs_perag_put(pag); |
2a82b8be8
|
162 163 164 165 166 167 168 |
*agp = ag; break; } /* Drop the reference on this AG, it's not usable. */ xfs_filestream_put_ag(mp, ag); next_ag: |
4196ac08c
|
169 |
xfs_perag_put(pag); |
2a82b8be8
|
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 |
/* Move to the next AG, wrapping to AG 0 if necessary. */ if (++ag >= mp->m_sb.sb_agcount) ag = 0; /* If a full pass of the AGs hasn't been done yet, continue. */ if (ag != startag) continue; /* Allow sleeping in xfs_alloc_pagf_init() on the 2nd pass. */ if (trylock != 0) { trylock = 0; continue; } /* Finally, if lowspace wasn't set, set it for the 3rd pass. */ if (!(flags & XFS_PICK_LOWSPACE)) { flags |= XFS_PICK_LOWSPACE; continue; } /* * Take the AG with the most free space, regardless of whether * it's already in use by another filestream. */ if (max_ag != NULLAGNUMBER) { xfs_filestream_get_ag(mp, max_ag); |
2a82b8be8
|
196 197 198 199 200 201 |
free = maxfree; *agp = max_ag; break; } /* take AG 0 if none matched */ |
b94acd478
|
202 |
trace_xfs_filestream_pick(ip, *agp, free, nscan); |
2a82b8be8
|
203 204 205 |
*agp = 0; return 0; } |
b94acd478
|
206 |
trace_xfs_filestream_pick(ip, *agp, free, nscan); |
2a82b8be8
|
207 |
|
2cd2ef6a3
|
208 |
if (*agp == NULLAGNUMBER) |
2a82b8be8
|
209 |
return 0; |
2a82b8be8
|
210 |
|
2451337dd
|
211 |
err = -ENOMEM; |
1919adda0
|
212 |
item = kmem_alloc(sizeof(*item), KM_MAYFAIL); |
2a82b8be8
|
213 |
if (!item) |
2cd2ef6a3
|
214 |
goto out_put_ag; |
2a82b8be8
|
215 |
|
2cd2ef6a3
|
216 |
item->ag = *agp; |
2a82b8be8
|
217 |
|
22328d712
|
218 |
err = xfs_mru_cache_insert(mp->m_filestream, ip->i_ino, &item->mru); |
2a82b8be8
|
219 |
if (err) { |
2451337dd
|
220 |
if (err == -EEXIST) |
2cd2ef6a3
|
221 222 |
err = 0; goto out_free_item; |
2a82b8be8
|
223 |
} |
2a82b8be8
|
224 |
return 0; |
2a82b8be8
|
225 |
|
2cd2ef6a3
|
226 |
out_free_item: |
1919adda0
|
227 |
kmem_free(item); |
2cd2ef6a3
|
228 229 230 |
out_put_ag: xfs_filestream_put_ag(mp, *agp); return err; |
2a82b8be8
|
231 |
} |
2cd2ef6a3
|
232 233 234 |
static struct xfs_inode * xfs_filestream_get_parent( struct xfs_inode *ip) |
2a82b8be8
|
235 |
{ |
2cd2ef6a3
|
236 237 |
struct inode *inode = VFS_I(ip), *dir = NULL; struct dentry *dentry, *parent; |
2a82b8be8
|
238 |
|
2cd2ef6a3
|
239 240 241 |
dentry = d_find_alias(inode); if (!dentry) goto out; |
2a82b8be8
|
242 |
|
2cd2ef6a3
|
243 244 245 |
parent = dget_parent(dentry); if (!parent) goto out_dput; |
2a82b8be8
|
246 |
|
2b0143b5c
|
247 |
dir = igrab(d_inode(parent)); |
2cd2ef6a3
|
248 |
dput(parent); |
2a82b8be8
|
249 |
|
2cd2ef6a3
|
250 251 252 253 |
out_dput: dput(dentry); out: return dir ? XFS_I(dir) : NULL; |
2a82b8be8
|
254 255 256 |
} /* |
3b8d90766
|
257 258 259 260 |
* Find the right allocation group for a file, either by finding an * existing file stream or creating a new one. * * Returns NULLAGNUMBER in case of an error. |
2a82b8be8
|
261 262 263 |
*/ xfs_agnumber_t xfs_filestream_lookup_ag( |
2cd2ef6a3
|
264 |
struct xfs_inode *ip) |
2a82b8be8
|
265 |
{ |
2cd2ef6a3
|
266 |
struct xfs_mount *mp = ip->i_mount; |
2cd2ef6a3
|
267 |
struct xfs_inode *pip = NULL; |
3b8d90766
|
268 |
xfs_agnumber_t startag, ag = NULLAGNUMBER; |
22328d712
|
269 |
struct xfs_mru_cache_elem *mru; |
2a82b8be8
|
270 |
|
c19b3b05a
|
271 |
ASSERT(S_ISREG(VFS_I(ip)->i_mode)); |
2a82b8be8
|
272 |
|
2cd2ef6a3
|
273 274 |
pip = xfs_filestream_get_parent(ip); if (!pip) |
b26384dc5
|
275 |
return NULLAGNUMBER; |
2cd2ef6a3
|
276 277 |
mru = xfs_mru_cache_lookup(mp->m_filestream, pip->i_ino); |
22328d712
|
278 |
if (mru) { |
3b8d90766
|
279 |
ag = container_of(mru, struct xfs_fstrm_item, mru)->ag; |
22328d712
|
280 |
xfs_mru_cache_done(mp->m_filestream); |
3b8d90766
|
281 |
|
7fcd3efa1
|
282 |
trace_xfs_filestream_lookup(mp, ip->i_ino, ag); |
3b8d90766
|
283 |
goto out; |
2a82b8be8
|
284 285 286 287 288 289 290 |
} /* * Set the starting AG using the rotor for inode32, otherwise * use the directory inode's AG. */ if (mp->m_flags & XFS_MOUNT_32BITINODES) { |
2cd2ef6a3
|
291 |
xfs_agnumber_t rotorstep = xfs_rotorstep; |
2a82b8be8
|
292 293 294 295 296 |
startag = (mp->m_agfrotor / rotorstep) % mp->m_sb.sb_agcount; mp->m_agfrotor = (mp->m_agfrotor + 1) % (mp->m_sb.sb_agcount * rotorstep); } else startag = XFS_INO_TO_AGNO(mp, pip->i_ino); |
3b8d90766
|
297 298 299 |
if (xfs_filestream_pick_ag(pip, startag, &ag, 0, 0)) ag = NULLAGNUMBER; out: |
44a8736bd
|
300 |
xfs_irele(pip); |
3b8d90766
|
301 |
return ag; |
2a82b8be8
|
302 303 304 |
} /* |
2cd2ef6a3
|
305 306 307 308 |
* Pick a new allocation group for the current file and its file stream. * * This is called when the allocator can't find a suitable extent in the * current AG, and we have to move the stream into a new AG with more space. |
2a82b8be8
|
309 310 311 |
*/ int xfs_filestream_new_ag( |
689881145
|
312 313 |
struct xfs_bmalloca *ap, xfs_agnumber_t *agp) |
2a82b8be8
|
314 |
{ |
2cd2ef6a3
|
315 316 317 318 |
struct xfs_inode *ip = ap->ip, *pip; struct xfs_mount *mp = ip->i_mount; xfs_extlen_t minlen = ap->length; xfs_agnumber_t startag = 0; |
292378edc
|
319 320 |
int flags = 0; int err = 0; |
2cd2ef6a3
|
321 |
struct xfs_mru_cache_elem *mru; |
2a82b8be8
|
322 |
|
2cd2ef6a3
|
323 |
*agp = NULLAGNUMBER; |
2a82b8be8
|
324 |
|
2cd2ef6a3
|
325 326 327 |
pip = xfs_filestream_get_parent(ip); if (!pip) goto exit; |
2a82b8be8
|
328 |
|
2cd2ef6a3
|
329 330 331 332 333 |
mru = xfs_mru_cache_remove(mp->m_filestream, pip->i_ino); if (mru) { struct xfs_fstrm_item *item = container_of(mru, struct xfs_fstrm_item, mru); startag = (item->ag + 1) % mp->m_sb.sb_agcount; |
2a82b8be8
|
334 |
} |
c34d570d1
|
335 |
if (ap->datatype & XFS_ALLOC_USERDATA) |
292378edc
|
336 |
flags |= XFS_PICK_USERDATA; |
1214f1cf6
|
337 |
if (ap->tp->t_flags & XFS_TRANS_LOWMODE) |
292378edc
|
338 |
flags |= XFS_PICK_LOWSPACE; |
2a82b8be8
|
339 |
|
2cd2ef6a3
|
340 |
err = xfs_filestream_pick_ag(pip, startag, agp, flags, minlen); |
2a82b8be8
|
341 342 |
/* |
2cd2ef6a3
|
343 |
* Only free the item here so we skip over the old AG earlier. |
2a82b8be8
|
344 |
*/ |
2cd2ef6a3
|
345 |
if (mru) |
7fcd3efa1
|
346 |
xfs_fstrm_free_func(mp, mru); |
2a82b8be8
|
347 |
|
44a8736bd
|
348 |
xfs_irele(pip); |
2cd2ef6a3
|
349 350 351 352 353 |
exit: if (*agp == NULLAGNUMBER) *agp = 0; return err; } |
2a82b8be8
|
354 |
|
2cd2ef6a3
|
355 356 357 358 359 360 |
void xfs_filestream_deassociate( struct xfs_inode *ip) { xfs_mru_cache_delete(ip->i_mount->m_filestream, ip->i_ino); } |
2a82b8be8
|
361 |
|
2cd2ef6a3
|
362 363 364 365 |
int xfs_filestream_mount( xfs_mount_t *mp) { |
2a82b8be8
|
366 |
/* |
2cd2ef6a3
|
367 368 369 370 371 |
* The filestream timer tunable is currently fixed within the range of * one second to four minutes, with five seconds being the default. The * group count is somewhat arbitrary, but it'd be nice to adhere to the * timer tunable to within about 10 percent. This requires at least 10 * groups. |
2a82b8be8
|
372 |
*/ |
7fcd3efa1
|
373 374 |
return xfs_mru_cache_create(&mp->m_filestream, mp, xfs_fstrm_centisecs * 10, 10, xfs_fstrm_free_func); |
2cd2ef6a3
|
375 |
} |
2a82b8be8
|
376 |
|
2cd2ef6a3
|
377 378 379 380 381 382 |
void xfs_filestream_unmount( xfs_mount_t *mp) { xfs_mru_cache_destroy(mp->m_filestream); } |