Re: [PATCH] mm: memory: Introduce new vmf_insert_mixed_mkwrite

From: kbuild test robot
Date: Mon Apr 23 2018 - 05:21:42 EST

Next message: Andrzej Hajda: "Re: [PATCH] drm/bridge/sii8620: fix link error"
Previous message: Maxime Ripard: "Re: [PATCH] Revert "drm/sun4i: add lvds mode_valid function""
In reply to: Souptick Joarder: "Re: [PATCH] mm: memory: Introduce new vmf_insert_mixed_mkwrite"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Souptick,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on mmotm/master]
[also build test ERROR on v4.17-rc2 next-20180423]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url: https://github.com/0day-ci/linux/commits/Souptick-Joarder/mm-memory-Introduce-new-vmf_insert_mixed_mkwrite/20180423-095015
base: git://git.cmpxchg.org/linux-mmotm.git master
config: x86_64-federa-25 (attached as .config)
compiler: gcc-7 (Debian 7.3.0-16) 7.3.0
reproduce:
# save the attached .config to linux build tree
make ARCH=x86_64

All errors (new ones prefixed by >>):

fs/dax.c: In function 'dax_iomap_pte_fault':
>> fs/dax.c:1265:12: error: implicit declaration of function 'vm_insert_mixed_mkwrite'; did you mean 'vmf_insert_mixed_mkwrite'? [-Werror=implicit-function-declaration]
error = vm_insert_mixed_mkwrite(vma, vaddr, pfn);
^~~~~~~~~~~~~~~~~~~~~~~
vmf_insert_mixed_mkwrite
cc1: some warnings being treated as errors

vim +1265 fs/dax.c

aaa422c4c Dan Williams 2017-11-13 1134
9a0dd4225 Jan Kara 2017-11-01 1135 static int dax_iomap_pte_fault(struct vm_fault *vmf, pfn_t *pfnp,
c0b246259 Jan Kara 2018-01-07 1136 int *iomap_errp, const struct iomap_ops *ops)
a7d73fe6c Christoph Hellwig 2016-09-19 1137 {
a0987ad5c Jan Kara 2017-11-01 1138 struct vm_area_struct *vma = vmf->vma;
a0987ad5c Jan Kara 2017-11-01 1139 struct address_space *mapping = vma->vm_file->f_mapping;
a7d73fe6c Christoph Hellwig 2016-09-19 1140 struct inode *inode = mapping->host;
1a29d85eb Jan Kara 2016-12-14 1141 unsigned long vaddr = vmf->address;
a7d73fe6c Christoph Hellwig 2016-09-19 1142 loff_t pos = (loff_t)vmf->pgoff << PAGE_SHIFT;
a7d73fe6c Christoph Hellwig 2016-09-19 1143 struct iomap iomap = { 0 };
9484ab1bf Jan Kara 2016-11-10 1144 unsigned flags = IOMAP_FAULT;
a7d73fe6c Christoph Hellwig 2016-09-19 1145 int error, major = 0;
d2c43ef13 Jan Kara 2017-11-01 1146 bool write = vmf->flags & FAULT_FLAG_WRITE;
caa51d26f Jan Kara 2017-11-01 1147 bool sync;
b1aa812b2 Jan Kara 2016-12-14 1148 int vmf_ret = 0;
a7d73fe6c Christoph Hellwig 2016-09-19 1149 void *entry;
1b5a1cb21 Jan Kara 2017-11-01 1150 pfn_t pfn;
a7d73fe6c Christoph Hellwig 2016-09-19 1151
a9c42b33e Ross Zwisler 2017-05-08 1152 trace_dax_pte_fault(inode, vmf, vmf_ret);
a7d73fe6c Christoph Hellwig 2016-09-19 1153 /*
a7d73fe6c Christoph Hellwig 2016-09-19 1154 * Check whether offset isn't beyond end of file now. Caller is supposed
a7d73fe6c Christoph Hellwig 2016-09-19 1155 * to hold locks serializing us with truncate / punch hole so this is
a7d73fe6c Christoph Hellwig 2016-09-19 1156 * a reliable test.
a7d73fe6c Christoph Hellwig 2016-09-19 1157 */
a9c42b33e Ross Zwisler 2017-05-08 1158 if (pos >= i_size_read(inode)) {
a9c42b33e Ross Zwisler 2017-05-08 1159 vmf_ret = VM_FAULT_SIGBUS;
a9c42b33e Ross Zwisler 2017-05-08 1160 goto out;
a9c42b33e Ross Zwisler 2017-05-08 1161 }
a7d73fe6c Christoph Hellwig 2016-09-19 1162
d2c43ef13 Jan Kara 2017-11-01 1163 if (write && !vmf->cow_page)
a7d73fe6c Christoph Hellwig 2016-09-19 1164 flags |= IOMAP_WRITE;
a7d73fe6c Christoph Hellwig 2016-09-19 1165
13e451fdc Jan Kara 2017-05-12 1166 entry = grab_mapping_entry(mapping, vmf->pgoff, 0);
13e451fdc Jan Kara 2017-05-12 1167 if (IS_ERR(entry)) {
13e451fdc Jan Kara 2017-05-12 1168 vmf_ret = dax_fault_return(PTR_ERR(entry));
13e451fdc Jan Kara 2017-05-12 1169 goto out;
13e451fdc Jan Kara 2017-05-12 1170 }
13e451fdc Jan Kara 2017-05-12 1171
a7d73fe6c Christoph Hellwig 2016-09-19 1172 /*
e2093926a Ross Zwisler 2017-06-02 1173 * It is possible, particularly with mixed reads & writes to private
e2093926a Ross Zwisler 2017-06-02 1174 * mappings, that we have raced with a PMD fault that overlaps with
e2093926a Ross Zwisler 2017-06-02 1175 * the PTE we need to set up. If so just return and the fault will be
e2093926a Ross Zwisler 2017-06-02 1176 * retried.
e2093926a Ross Zwisler 2017-06-02 1177 */
e2093926a Ross Zwisler 2017-06-02 1178 if (pmd_trans_huge(*vmf->pmd) || pmd_devmap(*vmf->pmd)) {
e2093926a Ross Zwisler 2017-06-02 1179 vmf_ret = VM_FAULT_NOPAGE;
e2093926a Ross Zwisler 2017-06-02 1180 goto unlock_entry;
e2093926a Ross Zwisler 2017-06-02 1181 }
e2093926a Ross Zwisler 2017-06-02 1182
e2093926a Ross Zwisler 2017-06-02 1183 /*
a7d73fe6c Christoph Hellwig 2016-09-19 1184 * Note that we don't bother to use iomap_apply here: DAX required
a7d73fe6c Christoph Hellwig 2016-09-19 1185 * the file system block size to be equal the page size, which means
a7d73fe6c Christoph Hellwig 2016-09-19 1186 * that we never have to deal with more than a single extent here.
a7d73fe6c Christoph Hellwig 2016-09-19 1187 */
a7d73fe6c Christoph Hellwig 2016-09-19 1188 error = ops->iomap_begin(inode, pos, PAGE_SIZE, flags, &iomap);
c0b246259 Jan Kara 2018-01-07 1189 if (iomap_errp)
c0b246259 Jan Kara 2018-01-07 1190 *iomap_errp = error;
a9c42b33e Ross Zwisler 2017-05-08 1191 if (error) {
a9c42b33e Ross Zwisler 2017-05-08 1192 vmf_ret = dax_fault_return(error);
13e451fdc Jan Kara 2017-05-12 1193 goto unlock_entry;
a9c42b33e Ross Zwisler 2017-05-08 1194 }
a7d73fe6c Christoph Hellwig 2016-09-19 1195 if (WARN_ON_ONCE(iomap.offset + iomap.length < pos + PAGE_SIZE)) {
13e451fdc Jan Kara 2017-05-12 1196 error = -EIO; /* fs corruption? */
13e451fdc Jan Kara 2017-05-12 1197 goto error_finish_iomap;
a7d73fe6c Christoph Hellwig 2016-09-19 1198 }
a7d73fe6c Christoph Hellwig 2016-09-19 1199
a7d73fe6c Christoph Hellwig 2016-09-19 1200 if (vmf->cow_page) {
31a6f1a6e Jan Kara 2017-11-01 1201 sector_t sector = dax_iomap_sector(&iomap, pos);
31a6f1a6e Jan Kara 2017-11-01 1202
a7d73fe6c Christoph Hellwig 2016-09-19 1203 switch (iomap.type) {
a7d73fe6c Christoph Hellwig 2016-09-19 1204 case IOMAP_HOLE:
a7d73fe6c Christoph Hellwig 2016-09-19 1205 case IOMAP_UNWRITTEN:
a7d73fe6c Christoph Hellwig 2016-09-19 1206 clear_user_highpage(vmf->cow_page, vaddr);
a7d73fe6c Christoph Hellwig 2016-09-19 1207 break;
a7d73fe6c Christoph Hellwig 2016-09-19 1208 case IOMAP_MAPPED:
cccbce671 Dan Williams 2017-01-27 1209 error = copy_user_dax(iomap.bdev, iomap.dax_dev,
cccbce671 Dan Williams 2017-01-27 1210 sector, PAGE_SIZE, vmf->cow_page, vaddr);
a7d73fe6c Christoph Hellwig 2016-09-19 1211 break;
a7d73fe6c Christoph Hellwig 2016-09-19 1212 default:
a7d73fe6c Christoph Hellwig 2016-09-19 1213 WARN_ON_ONCE(1);
a7d73fe6c Christoph Hellwig 2016-09-19 1214 error = -EIO;
a7d73fe6c Christoph Hellwig 2016-09-19 1215 break;
a7d73fe6c Christoph Hellwig 2016-09-19 1216 }
a7d73fe6c Christoph Hellwig 2016-09-19 1217
a7d73fe6c Christoph Hellwig 2016-09-19 1218 if (error)
13e451fdc Jan Kara 2017-05-12 1219 goto error_finish_iomap;
b1aa812b2 Jan Kara 2016-12-14 1220
b1aa812b2 Jan Kara 2016-12-14 1221 __SetPageUptodate(vmf->cow_page);
b1aa812b2 Jan Kara 2016-12-14 1222 vmf_ret = finish_fault(vmf);
b1aa812b2 Jan Kara 2016-12-14 1223 if (!vmf_ret)
b1aa812b2 Jan Kara 2016-12-14 1224 vmf_ret = VM_FAULT_DONE_COW;
13e451fdc Jan Kara 2017-05-12 1225 goto finish_iomap;
a7d73fe6c Christoph Hellwig 2016-09-19 1226 }
a7d73fe6c Christoph Hellwig 2016-09-19 1227
aaa422c4c Dan Williams 2017-11-13 1228 sync = dax_fault_is_synchronous(flags, vma, &iomap);
caa51d26f Jan Kara 2017-11-01 1229
a7d73fe6c Christoph Hellwig 2016-09-19 1230 switch (iomap.type) {
a7d73fe6c Christoph Hellwig 2016-09-19 1231 case IOMAP_MAPPED:
a7d73fe6c Christoph Hellwig 2016-09-19 1232 if (iomap.flags & IOMAP_F_NEW) {
a7d73fe6c Christoph Hellwig 2016-09-19 1233 count_vm_event(PGMAJFAULT);
a0987ad5c Jan Kara 2017-11-01 1234 count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
a7d73fe6c Christoph Hellwig 2016-09-19 1235 major = VM_FAULT_MAJOR;
a7d73fe6c Christoph Hellwig 2016-09-19 1236 }
1b5a1cb21 Jan Kara 2017-11-01 1237 error = dax_iomap_pfn(&iomap, pos, PAGE_SIZE, &pfn);
1b5a1cb21 Jan Kara 2017-11-01 1238 if (error < 0)
1b5a1cb21 Jan Kara 2017-11-01 1239 goto error_finish_iomap;
1b5a1cb21 Jan Kara 2017-11-01 1240
56addbc73 Andrew Morton 2018-04-14 1241 entry = dax_insert_mapping_entry(mapping, vmf, entry, pfn,
caa51d26f Jan Kara 2017-11-01 1242 0, write && !sync);
1b5a1cb21 Jan Kara 2017-11-01 1243 if (IS_ERR(entry)) {
1b5a1cb21 Jan Kara 2017-11-01 1244 error = PTR_ERR(entry);
1b5a1cb21 Jan Kara 2017-11-01 1245 goto error_finish_iomap;
1b5a1cb21 Jan Kara 2017-11-01 1246 }
1b5a1cb21 Jan Kara 2017-11-01 1247
caa51d26f Jan Kara 2017-11-01 1248 /*
caa51d26f Jan Kara 2017-11-01 1249 * If we are doing synchronous page fault and inode needs fsync,
caa51d26f Jan Kara 2017-11-01 1250 * we can insert PTE into page tables only after that happens.
caa51d26f Jan Kara 2017-11-01 1251 * Skip insertion for now and return the pfn so that caller can
caa51d26f Jan Kara 2017-11-01 1252 * insert it after fsync is done.
caa51d26f Jan Kara 2017-11-01 1253 */
caa51d26f Jan Kara 2017-11-01 1254 if (sync) {
caa51d26f Jan Kara 2017-11-01 1255 if (WARN_ON_ONCE(!pfnp)) {
caa51d26f Jan Kara 2017-11-01 1256 error = -EIO;
caa51d26f Jan Kara 2017-11-01 1257 goto error_finish_iomap;
caa51d26f Jan Kara 2017-11-01 1258 }
caa51d26f Jan Kara 2017-11-01 1259 *pfnp = pfn;
caa51d26f Jan Kara 2017-11-01 1260 vmf_ret = VM_FAULT_NEEDDSYNC | major;
caa51d26f Jan Kara 2017-11-01 1261 goto finish_iomap;
caa51d26f Jan Kara 2017-11-01 1262 }
1b5a1cb21 Jan Kara 2017-11-01 1263 trace_dax_insert_mapping(inode, vmf, entry);
1b5a1cb21 Jan Kara 2017-11-01 1264 if (write)
1b5a1cb21 Jan Kara 2017-11-01 @1265 error = vm_insert_mixed_mkwrite(vma, vaddr, pfn);
1b5a1cb21 Jan Kara 2017-11-01 1266 else
1b5a1cb21 Jan Kara 2017-11-01 1267 error = vm_insert_mixed(vma, vaddr, pfn);
1b5a1cb21 Jan Kara 2017-11-01 1268
9f141d6ef Jan Kara 2016-10-19 1269 /* -EBUSY is fine, somebody else faulted on the same PTE */
9f141d6ef Jan Kara 2016-10-19 1270 if (error == -EBUSY)
9f141d6ef Jan Kara 2016-10-19 1271 error = 0;
a7d73fe6c Christoph Hellwig 2016-09-19 1272 break;
a7d73fe6c Christoph Hellwig 2016-09-19 1273 case IOMAP_UNWRITTEN:
a7d73fe6c Christoph Hellwig 2016-09-19 1274 case IOMAP_HOLE:
d2c43ef13 Jan Kara 2017-11-01 1275 if (!write) {
91d25ba8a Ross Zwisler 2017-09-06 1276 vmf_ret = dax_load_hole(mapping, entry, vmf);
13e451fdc Jan Kara 2017-05-12 1277 goto finish_iomap;
1550290b0 Ross Zwisler 2016-11-08 1278 }
a7d73fe6c Christoph Hellwig 2016-09-19 1279 /*FALLTHRU*/
a7d73fe6c Christoph Hellwig 2016-09-19 1280 default:
a7d73fe6c Christoph Hellwig 2016-09-19 1281 WARN_ON_ONCE(1);
a7d73fe6c Christoph Hellwig 2016-09-19 1282 error = -EIO;
a7d73fe6c Christoph Hellwig 2016-09-19 1283 break;
a7d73fe6c Christoph Hellwig 2016-09-19 1284 }
a7d73fe6c Christoph Hellwig 2016-09-19 1285
13e451fdc Jan Kara 2017-05-12 1286 error_finish_iomap:
9f141d6ef Jan Kara 2016-10-19 1287 vmf_ret = dax_fault_return(error) | major;
1550290b0 Ross Zwisler 2016-11-08 1288 finish_iomap:
1550290b0 Ross Zwisler 2016-11-08 1289 if (ops->iomap_end) {
9f141d6ef Jan Kara 2016-10-19 1290 int copied = PAGE_SIZE;
9f141d6ef Jan Kara 2016-10-19 1291
9f141d6ef Jan Kara 2016-10-19 1292 if (vmf_ret & VM_FAULT_ERROR)
9f141d6ef Jan Kara 2016-10-19 1293 copied = 0;
9f141d6ef Jan Kara 2016-10-19 1294 /*
9f141d6ef Jan Kara 2016-10-19 1295 * The fault is done by now and there's no way back (other
9f141d6ef Jan Kara 2016-10-19 1296 * thread may be already happily using PTE we have installed).
9f141d6ef Jan Kara 2016-10-19 1297 * Just ignore error from ->iomap_end since we cannot do much
9f141d6ef Jan Kara 2016-10-19 1298 * with it.
9f141d6ef Jan Kara 2016-10-19 1299 */
9f141d6ef Jan Kara 2016-10-19 1300 ops->iomap_end(inode, pos, PAGE_SIZE, copied, flags, &iomap);
1550290b0 Ross Zwisler 2016-11-08 1301 }
13e451fdc Jan Kara 2017-05-12 1302 unlock_entry:
91d25ba8a Ross Zwisler 2017-09-06 1303 put_locked_mapping_entry(mapping, vmf->pgoff);
a9c42b33e Ross Zwisler 2017-05-08 1304 out:
a9c42b33e Ross Zwisler 2017-05-08 1305 trace_dax_pte_fault_done(inode, vmf, vmf_ret);
b1aa812b2 Jan Kara 2016-12-14 1306 return vmf_ret;
1550290b0 Ross Zwisler 2016-11-08 1307 }
642261ac9 Ross Zwisler 2016-11-08 1308

:::::: The code at line 1265 was first introduced by commit
:::::: 1b5a1cb21e0cdfb001050c76dc31039cdece1a63 dax: Inline dax_insert_mapping() into the callsite

:::::: TO: Jan Kara <jack@xxxxxxx>
:::::: CC: Dan Williams <dan.j.williams@xxxxxxxxx>

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

Attachment: .config.gz
Description: application/gzip

Next message: Andrzej Hajda: "Re: [PATCH] drm/bridge/sii8620: fix link error"
Previous message: Maxime Ripard: "Re: [PATCH] Revert "drm/sun4i: add lvds mode_valid function""
In reply to: Souptick Joarder: "Re: [PATCH] mm: memory: Introduce new vmf_insert_mixed_mkwrite"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]