Hi, Mauro.
> > +
> > +#define pr_fmt(fmt) KBUILD_MODNAME ": %s: " fmt, __func__
>
> It would be better to use dev_foo() debug macros instead of
> pr_foo() ones.
I got comment for this previous version patch as below
--------------------------------------------------------------------------------------
The best would be to se dev_err() & friends for printing messages,
as they print the device's name as filled at struct device.
If you don't use, please add a define that will print the name at the logs,
like:
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
either at the begining of the driver or at some header file.
Btw, I'm noticing that you're also using dev_err() on other places of the code.
Please standardize. OK, on a few places, you may still need to use pr_err(),
if you need to print a message before initializing struct device, but I suspect
that you can init
--------------------------------------------------------------------------------------
You pointed out here before. Because dev_foo () and pr_foo () were mixed.
We standardize with pr_foo() because the logs is outputted before getting the
device structure.
Is it better to use dev_foo() where we can use it?
> > +static int cxd2880_stop_feed(struct dvb_demux_feed *feed) {
> > + int i = 0;
> > + int ret;
> > + struct dvb_demux *demux = NULL;
> > + struct cxd2880_dvb_spi *dvb_spi = NULL;
> > +
> > + if (!feed) {
> > + pr_err("invalid arg\n");
> > + return -EINVAL;
> > + }
> > +
> > + demux = feed->demux;
> > + if (!demux) {
> > + pr_err("feed->demux is NULL\n");
> > + return -EINVAL;
> > + }
> > + dvb_spi = demux->priv;
> > +
> > + if (!dvb_spi->feed_count) {
> > + pr_err("no feed is started\n");
> > + return -EINVAL;
> > + }
> > +
> > + if (feed->pid == 0x2000) {
> > + /*
> > + * Special PID case.
> > + * Number of 0x2000 feed request was stored
> > + * in dvb_spi->all_pid_feed_count.
> > + */
> > + if (dvb_spi->all_pid_feed_count <= 0) {
> > + pr_err("PID %d not found.\n", feed->pid);
> > + return -EINVAL;
> > + }
> > + dvb_spi->all_pid_feed_count--;
> > + } else {
> > + struct cxd2880_pid_filter_config cfgtmp;
> > +
> > + cfgtmp = dvb_spi->filter_config;
> > +
> > + for (i = 0; i < CXD2880_MAX_FILTER_SIZE; i++) {
> > + if (feed->pid == cfgtmp.pid_config[i].pid &&
> > + cfgtmp.pid_config[i].is_enable != 0) {
> > + cfgtmp.pid_config[i].is_enable = 0;
> > + cfgtmp.pid_config[i].pid = 0;
> > + pr_debug("removed PID %d from #%d\n",
> > + feed->pid, i);
> > + break;
> > + }
> > + }
> > + dvb_spi->filter_config = cfgtmp;
> > +
> > + if (i == CXD2880_MAX_FILTER_SIZE) {
> > + pr_err("PID %d not found\n", feed->pid);
> > + return -EINVAL;
> > + }
> > + }
> > +
> > + ret = cxd2880_update_pid_filter(dvb_spi,
> > + &dvb_spi->filter_config,
> > + dvb_spi->all_pid_feed_count >
> 0);
> > + dvb_spi->feed_count--;
> > +
> > + if (dvb_spi->feed_count == 0) {
> > + int ret_stop = 0;
> > +
> > + ret_stop =
> kthread_stop(dvb_spi->cxd2880_ts_read_thread);
> > + if (ret_stop) {
> > + pr_err("'kthread_stop failed. (%d)\n",
> ret_stop);
> > + ret = ret_stop;
> > + }
> > + kfree(dvb_spi->ts_buf);
> > + dvb_spi->ts_buf = NULL;
> > + }
> > +
> > + pr_debug("stop feed ok.(count %d)\n", dvb_spi->feed_count);
> > +
> > + return ret;
> > +}
>
> I have the feeling that I've seen the code above before at the dvb core.
> Any reason why don't use the already-existing code at dvb_demux.c &
> friends?
The CXD2880 HW PID filter is used to decrease the amount of TS data transferred
via limited bit rate SPI interface.
Thanks,
Takiguchi
