Hey
On Wed, Aug 3, 2016 at 9:01 PM, Chris Wilson chris@chris-wilson.co.uk wrote:
On Wed, Aug 03, 2016 at 08:04:26PM +0200, David Herrmann wrote:
@@ -98,13 +99,14 @@ int drm_clients_info(struct seq_file *m, void *data)
rcu_read_lock(); /* locks pid_task()->comm */ task = pid_task(priv->pid, PIDTYPE_PID);
uid = priv->filp ? priv->filp->f_cred->euid : GLOBAL_ROOT_UID; seq_printf(m, "%20s %5d %3d %c %c %5d %10u\n", task ? task->comm : "<unknown>", pid_vnr(priv->pid), priv->minor->index, drm_is_current_master(priv) ? 'y' : 'n', priv->authenticated ? 'y' : 'n',
from_kuid_munged(seq_user_ns(m), priv->uid),
from_kuid_munged(seq_user_ns(m), uid), priv->magic); rcu_read_unlock(); }diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c index 33af4a5..49cd835 100644 --- a/drivers/gpu/drm/drm_ioctl.c +++ b/drivers/gpu/drm/drm_ioctl.c @@ -191,7 +191,9 @@ static int drm_getclient(struct drm_device *dev, void *data, client->auth = file_priv->authenticated; client->pid = pid_vnr(file_priv->pid); client->uid = from_kuid_munged(current_user_ns(),
file_priv->uid);
file_priv->filp ?file_priv->filp->f_cred->euid :GLOBAL_ROOT_UID);Why can't we use task_euid(pid_task(file_priv->pid)) here as well?
task_euid() changes semantics. With this patch I just tried to get rid of "filp" usage, but keep semantics, so the ABI does not suddently change.
Note that both calls are actually changed in follow-ups. "uid" is just cleared to "overflowuid", since no-one ever looks at that field. And "pid" is set to "pid_vnr(current)". However, I explicitly split those patches to make sure it is easier to bisect.
Thanks David