[media] Remove the double symbol increment hack from drxk_hard
Both ngene and ddbrige calls dvb_attach once for drxk_attach.
The logic used there, and by tda18271c2dd driver is different
from similar logic on other frontends.
The right fix is to change them to use the same logic, but,
while we don't do that, we need to patch em28xx-dvb in order
to do cope with ngene/ddbridge magic.
While here, document why drxk_t_release should do nothing.
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
diff --git a/drivers/media/dvb/frontends/drxk_hard.c b/drivers/media/dvb/frontends/drxk_hard.c
index d503558..a0e2ff5 100644
--- a/drivers/media/dvb/frontends/drxk_hard.c
+++ b/drivers/media/dvb/frontends/drxk_hard.c
@@ -6313,12 +6313,10 @@
static void drxk_t_release(struct dvb_frontend *fe)
{
-#if 0
- struct drxk_state *state = fe->demodulator_priv;
-
- dprintk(1, "\n");
- kfree(state);
-#endif
+ /*
+ * There's nothing to release here, as the state struct
+ * is already freed by drxk_c_release.
+ */
}
static int drxk_t_init(struct dvb_frontend *fe)
@@ -6451,17 +6449,6 @@
goto error;
*fe_t = &state->t_frontend;
-#ifdef CONFIG_MEDIA_ATTACH
- /*
- * HACK: As this function initializes both DVB-T and DVB-C fe symbols,
- * and calling it twice would create the state twice, leading into
- * memory leaks, the right way is to call it only once. However, dvb
- * release functions will call symbol_put twice. So, the solution is to
- * artificially increment the usage count, in order to allow the
- * driver to be released.
- */
- symbol_get(drxk_attach);
-#endif
return &state->c_frontend;
error: