| Message ID | 20210730052355.77289-1-someguy@effective-light.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | apparmor: use per file locks for transactional queries | expand |
On 7/29/21 10:23 PM, Hamza Mahfooz wrote: > As made mention of in commit 1dea3b41e84c5 ("apparmor: speed up > transactional queries"), a single lock is currently used to synchronize > transactional queries. We can, use the lock allocated for each file by > VFS instead. > > Signed-off-by: Hamza Mahfooz <someguy@effective-light.com> Acked-by: John Johansen <john.johansen@canonical.com> I'll pull this into my tree asap > --- > security/apparmor/apparmorfs.c | 11 +++++------ > 1 file changed, 5 insertions(+), 6 deletions(-) > > diff --git a/security/apparmor/apparmorfs.c b/security/apparmor/apparmorfs.c > index 2ee3b3d29f10..c0b626a271a0 100644 > --- a/security/apparmor/apparmorfs.c > +++ b/security/apparmor/apparmorfs.c > @@ -812,8 +812,6 @@ struct multi_transaction { > }; > > #define MULTI_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct multi_transaction)) > -/* TODO: replace with per file lock */ > -static DEFINE_SPINLOCK(multi_transaction_lock); > > static void multi_transaction_kref(struct kref *kref) > { > @@ -847,10 +845,10 @@ static void multi_transaction_set(struct file *file, > AA_BUG(n > MULTI_TRANSACTION_LIMIT); > > new->size = n; > - spin_lock(&multi_transaction_lock); > + spin_lock(&file->f_lock); > old = (struct multi_transaction *) file->private_data; > file->private_data = new; > - spin_unlock(&multi_transaction_lock); > + spin_unlock(&file->f_lock); > put_multi_transaction(old); > } > > @@ -879,9 +877,10 @@ static ssize_t multi_transaction_read(struct file *file, char __user *buf, > struct multi_transaction *t; > ssize_t ret; > > - spin_lock(&multi_transaction_lock); > + spin_lock(&file->f_lock); > t = get_multi_transaction(file->private_data); > - spin_unlock(&multi_transaction_lock); > + spin_unlock(&file->f_lock); > + > if (!t) > return 0; > >
diff --git a/security/apparmor/apparmorfs.c b/security/apparmor/apparmorfs.c index 2ee3b3d29f10..c0b626a271a0 100644 --- a/security/apparmor/apparmorfs.c +++ b/security/apparmor/apparmorfs.c @@ -812,8 +812,6 @@ struct multi_transaction { }; #define MULTI_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct multi_transaction)) -/* TODO: replace with per file lock */ -static DEFINE_SPINLOCK(multi_transaction_lock); static void multi_transaction_kref(struct kref *kref) { @@ -847,10 +845,10 @@ static void multi_transaction_set(struct file *file, AA_BUG(n > MULTI_TRANSACTION_LIMIT); new->size = n; - spin_lock(&multi_transaction_lock); + spin_lock(&file->f_lock); old = (struct multi_transaction *) file->private_data; file->private_data = new; - spin_unlock(&multi_transaction_lock); + spin_unlock(&file->f_lock); put_multi_transaction(old); } @@ -879,9 +877,10 @@ static ssize_t multi_transaction_read(struct file *file, char __user *buf, struct multi_transaction *t; ssize_t ret; - spin_lock(&multi_transaction_lock); + spin_lock(&file->f_lock); t = get_multi_transaction(file->private_data); - spin_unlock(&multi_transaction_lock); + spin_unlock(&file->f_lock); + if (!t) return 0;
As made mention of in commit 1dea3b41e84c5 ("apparmor: speed up transactional queries"), a single lock is currently used to synchronize transactional queries. We can, use the lock allocated for each file by VFS instead. Signed-off-by: Hamza Mahfooz <someguy@effective-light.com> --- security/apparmor/apparmorfs.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)