Revision 6e2df0581f569038719cf2bc2b3baa3fcc83cab4 authored by Peter Zijlstra on 08 November 2019, 10:11:52 UTC, committed by Peter Zijlstra on 08 November 2019, 21:34:14 UTC
Commit 67692435c411 ("sched: Rework pick_next_task() slow-path")
inadvertly introduced a race because it changed a previously
unexplored dependency between dropping the rq->lock and
sched_class::put_prev_task().
The comments about dropping rq->lock, in for example
newidle_balance(), only mentions the task being current and ->on_cpu
being set. But when we look at the 'change' pattern (in for example
sched_setnuma()):
queued = task_on_rq_queued(p); /* p->on_rq == TASK_ON_RQ_QUEUED */
running = task_current(rq, p); /* rq->curr == p */
if (queued)
dequeue_task(...);
if (running)
put_prev_task(...);
/* change task properties */
if (queued)
enqueue_task(...);
if (running)
set_next_task(...);
It becomes obvious that if we do this after put_prev_task() has
already been called on @p, things go sideways. This is exactly what
the commit in question allows to happen when it does:
prev->sched_class->put_prev_task(rq, prev, rf);
if (!rq->nr_running)
newidle_balance(rq, rf);
The newidle_balance() call will drop rq->lock after we've called
put_prev_task() and that allows the above 'change' pattern to
interleave and mess up the state.
Furthermore, it turns out we lost the RT-pull when we put the last DL
task.
Fix both problems by extracting the balancing from put_prev_task() and
doing a multi-class balance() pass before put_prev_task().
Fixes: 67692435c411 ("sched: Rework pick_next_task() slow-path")
Reported-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Quentin Perret <qperret@google.com>
Tested-by: Valentin Schneider <valentin.schneider@arm.com>
1 parent e3b8b6a
acl.c
/*
* JFFS2 -- Journalling Flash File System, Version 2.
*
* Copyright © 2006 NEC Corporation
*
* Created by KaiGai Kohei <kaigai@ak.jp.nec.com>
*
* For licensing information, see the file 'LICENCE' in this directory.
*
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/sched.h>
#include <linux/time.h>
#include <linux/crc32.h>
#include <linux/jffs2.h>
#include <linux/xattr.h>
#include <linux/posix_acl_xattr.h>
#include <linux/mtd/mtd.h>
#include "nodelist.h"
static size_t jffs2_acl_size(int count)
{
if (count <= 4) {
return sizeof(struct jffs2_acl_header)
+ count * sizeof(struct jffs2_acl_entry_short);
} else {
return sizeof(struct jffs2_acl_header)
+ 4 * sizeof(struct jffs2_acl_entry_short)
+ (count - 4) * sizeof(struct jffs2_acl_entry);
}
}
static int jffs2_acl_count(size_t size)
{
size_t s;
size -= sizeof(struct jffs2_acl_header);
if (size < 4 * sizeof(struct jffs2_acl_entry_short)) {
if (size % sizeof(struct jffs2_acl_entry_short))
return -1;
return size / sizeof(struct jffs2_acl_entry_short);
} else {
s = size - 4 * sizeof(struct jffs2_acl_entry_short);
if (s % sizeof(struct jffs2_acl_entry))
return -1;
return s / sizeof(struct jffs2_acl_entry) + 4;
}
}
static struct posix_acl *jffs2_acl_from_medium(void *value, size_t size)
{
void *end = value + size;
struct jffs2_acl_header *header = value;
struct jffs2_acl_entry *entry;
struct posix_acl *acl;
uint32_t ver;
int i, count;
if (!value)
return NULL;
if (size < sizeof(struct jffs2_acl_header))
return ERR_PTR(-EINVAL);
ver = je32_to_cpu(header->a_version);
if (ver != JFFS2_ACL_VERSION) {
JFFS2_WARNING("Invalid ACL version. (=%u)\n", ver);
return ERR_PTR(-EINVAL);
}
value += sizeof(struct jffs2_acl_header);
count = jffs2_acl_count(size);
if (count < 0)
return ERR_PTR(-EINVAL);
if (count == 0)
return NULL;
acl = posix_acl_alloc(count, GFP_KERNEL);
if (!acl)
return ERR_PTR(-ENOMEM);
for (i=0; i < count; i++) {
entry = value;
if (value + sizeof(struct jffs2_acl_entry_short) > end)
goto fail;
acl->a_entries[i].e_tag = je16_to_cpu(entry->e_tag);
acl->a_entries[i].e_perm = je16_to_cpu(entry->e_perm);
switch (acl->a_entries[i].e_tag) {
case ACL_USER_OBJ:
case ACL_GROUP_OBJ:
case ACL_MASK:
case ACL_OTHER:
value += sizeof(struct jffs2_acl_entry_short);
break;
case ACL_USER:
value += sizeof(struct jffs2_acl_entry);
if (value > end)
goto fail;
acl->a_entries[i].e_uid =
make_kuid(&init_user_ns,
je32_to_cpu(entry->e_id));
break;
case ACL_GROUP:
value += sizeof(struct jffs2_acl_entry);
if (value > end)
goto fail;
acl->a_entries[i].e_gid =
make_kgid(&init_user_ns,
je32_to_cpu(entry->e_id));
break;
default:
goto fail;
}
}
if (value != end)
goto fail;
return acl;
fail:
posix_acl_release(acl);
return ERR_PTR(-EINVAL);
}
static void *jffs2_acl_to_medium(const struct posix_acl *acl, size_t *size)
{
struct jffs2_acl_header *header;
struct jffs2_acl_entry *entry;
void *e;
size_t i;
*size = jffs2_acl_size(acl->a_count);
header = kmalloc(struct_size(header, a_entries, acl->a_count),
GFP_KERNEL);
if (!header)
return ERR_PTR(-ENOMEM);
header->a_version = cpu_to_je32(JFFS2_ACL_VERSION);
e = header + 1;
for (i=0; i < acl->a_count; i++) {
const struct posix_acl_entry *acl_e = &acl->a_entries[i];
entry = e;
entry->e_tag = cpu_to_je16(acl_e->e_tag);
entry->e_perm = cpu_to_je16(acl_e->e_perm);
switch(acl_e->e_tag) {
case ACL_USER:
entry->e_id = cpu_to_je32(
from_kuid(&init_user_ns, acl_e->e_uid));
e += sizeof(struct jffs2_acl_entry);
break;
case ACL_GROUP:
entry->e_id = cpu_to_je32(
from_kgid(&init_user_ns, acl_e->e_gid));
e += sizeof(struct jffs2_acl_entry);
break;
case ACL_USER_OBJ:
case ACL_GROUP_OBJ:
case ACL_MASK:
case ACL_OTHER:
e += sizeof(struct jffs2_acl_entry_short);
break;
default:
goto fail;
}
}
return header;
fail:
kfree(header);
return ERR_PTR(-EINVAL);
}
struct posix_acl *jffs2_get_acl(struct inode *inode, int type)
{
struct posix_acl *acl;
char *value = NULL;
int rc, xprefix;
switch (type) {
case ACL_TYPE_ACCESS:
xprefix = JFFS2_XPREFIX_ACL_ACCESS;
break;
case ACL_TYPE_DEFAULT:
xprefix = JFFS2_XPREFIX_ACL_DEFAULT;
break;
default:
BUG();
}
rc = do_jffs2_getxattr(inode, xprefix, "", NULL, 0);
if (rc > 0) {
value = kmalloc(rc, GFP_KERNEL);
if (!value)
return ERR_PTR(-ENOMEM);
rc = do_jffs2_getxattr(inode, xprefix, "", value, rc);
}
if (rc > 0) {
acl = jffs2_acl_from_medium(value, rc);
} else if (rc == -ENODATA || rc == -ENOSYS) {
acl = NULL;
} else {
acl = ERR_PTR(rc);
}
kfree(value);
return acl;
}
static int __jffs2_set_acl(struct inode *inode, int xprefix, struct posix_acl *acl)
{
char *value = NULL;
size_t size = 0;
int rc;
if (acl) {
value = jffs2_acl_to_medium(acl, &size);
if (IS_ERR(value))
return PTR_ERR(value);
}
rc = do_jffs2_setxattr(inode, xprefix, "", value, size, 0);
if (!value && rc == -ENODATA)
rc = 0;
kfree(value);
return rc;
}
int jffs2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
{
int rc, xprefix;
switch (type) {
case ACL_TYPE_ACCESS:
xprefix = JFFS2_XPREFIX_ACL_ACCESS;
if (acl) {
umode_t mode;
rc = posix_acl_update_mode(inode, &mode, &acl);
if (rc)
return rc;
if (inode->i_mode != mode) {
struct iattr attr;
attr.ia_valid = ATTR_MODE | ATTR_CTIME;
attr.ia_mode = mode;
attr.ia_ctime = current_time(inode);
rc = jffs2_do_setattr(inode, &attr);
if (rc < 0)
return rc;
}
}
break;
case ACL_TYPE_DEFAULT:
xprefix = JFFS2_XPREFIX_ACL_DEFAULT;
if (!S_ISDIR(inode->i_mode))
return acl ? -EACCES : 0;
break;
default:
return -EINVAL;
}
rc = __jffs2_set_acl(inode, xprefix, acl);
if (!rc)
set_cached_acl(inode, type, acl);
return rc;
}
int jffs2_init_acl_pre(struct inode *dir_i, struct inode *inode, umode_t *i_mode)
{
struct posix_acl *default_acl, *acl;
int rc;
cache_no_acl(inode);
rc = posix_acl_create(dir_i, i_mode, &default_acl, &acl);
if (rc)
return rc;
if (default_acl) {
set_cached_acl(inode, ACL_TYPE_DEFAULT, default_acl);
posix_acl_release(default_acl);
}
if (acl) {
set_cached_acl(inode, ACL_TYPE_ACCESS, acl);
posix_acl_release(acl);
}
return 0;
}
int jffs2_init_acl_post(struct inode *inode)
{
int rc;
if (inode->i_default_acl) {
rc = __jffs2_set_acl(inode, JFFS2_XPREFIX_ACL_DEFAULT, inode->i_default_acl);
if (rc)
return rc;
}
if (inode->i_acl) {
rc = __jffs2_set_acl(inode, JFFS2_XPREFIX_ACL_ACCESS, inode->i_acl);
if (rc)
return rc;
}
return 0;
}
Computing file changes ...
