| Message ID | 20200701013320.130441-2-drew@beagleboard.org |
|---|---|
| State | New |
| Headers | show |
| Series | pinctrl: single: support #pinctrl-cells = 2 | expand |
On Tuesday, June 30, 2020 6:33:19 PM PDT Drew Fustini wrote: > If "pinctrl-single,pins" has 3 arguments (offset, conf, mux), then > pcs_parse_one_pinctrl_entry() does an OR operation on conf and mux to > get the value to store in the register. > - vals[found].val = pinctrl_spec.args[1]; > + > + switch (pinctrl_spec.args_count) { > + case 2: > + vals[found].val = pinctrl_spec.args[1]; > + break; > + case 3: > + vals[found].val = (pinctrl_spec.args[1] | pinctrl_spec.args[2]); > + break; > + } > > dev_dbg(pcs->dev, "%pOFn index: 0x%x value: 0x%x\n", > pinctrl_spec.np, offset, pinctrl_spec.args[1]); If #pinctrl-cells value is greater than 2, nothing will set vals[found].val to anything other than zero (from when it's calloc'ed) and the pinctrl will silently be programmed to zero. The debug printout was not change to print vals[found].val, so it will continue to print the value of the 2nd cell. The result is that a #pinctrl-cells of 3 will produce no warning or error, program the pinctrl to zero, whilst at the same time emit debug log messages that it is programming the expected values. The device tree documentation still states that #pinctrl-cells must be 1 when using pinctrl-single,pins. This new special case of ORing two values is not documented.
On Tue, Sep 08, 2020 at 04:52:58PM -0700, Trent Piepho wrote: > On Tuesday, June 30, 2020 6:33:19 PM PDT Drew Fustini wrote: > > If "pinctrl-single,pins" has 3 arguments (offset, conf, mux), then > > pcs_parse_one_pinctrl_entry() does an OR operation on conf and mux to > > get the value to store in the register. > > > > - vals[found].val = pinctrl_spec.args[1]; > > + > > + switch (pinctrl_spec.args_count) { > > + case 2: > > + vals[found].val = pinctrl_spec.args[1]; > > + break; > > + case 3: > > + vals[found].val = (pinctrl_spec.args[1] | > pinctrl_spec.args[2]); > > + break; > > + } > > > > dev_dbg(pcs->dev, "%pOFn index: 0x%x value: 0x%x\n", > > pinctrl_spec.np, offset, > pinctrl_spec.args[1]); > > If #pinctrl-cells value is greater than 2, nothing will set vals[found].val to > anything other than zero (from when it's calloc'ed) and the pinctrl will > silently be programmed to zero. If #pinctrl-cells is 3, then it will be: vals[found].val = (pinctrl_spec.args[1] | pinctrl_spec.args[2]); Do you mean if #pinctrl-cells is great than 3 then it will just have a default value of zero? That does appear to be the case and is probably not the behavior we want. Thank you for pointing this out. Earlier, there is a check to make sure there are at least 2 arguments: if (pinctrl_spec.args_count < 2) { dev_err(pcs->dev, "invalid args_count for spec: %i\n", pinctrl_spec.args_count); break; } I'll submit a patch where the upper bound is also checked: if (pinctrl_spec.args_count < 2 || pinctrl_spec.args_count > 3) { dev_err(pcs->dev, "invalid args_count for spec: %i\n", pinctrl_spec.args_count); break; } > The debug printout was not change to print vals[found].val, so it will > continue to print the value of the 2nd cell. Thank you for pointing this out. Yes, this is an oversight and I will submit a patch. > The result is that a #pinctrl-cells of 3 will produce no warning or error, > program the pinctrl to zero, whilst at the same time emit debug log messages > that it is programming the expected values. > > The device tree documentation still states that #pinctrl-cells must be 1 when > using pinctrl-single,pins. This new special case of ORing two values is not > documented. This is a good point, too. I will make a patch to update the documentation. -Drew
On Sun, Sep 13, 2020 at 09:42:33PM +0200, Drew Fustini wrote: > On Tue, Sep 08, 2020 at 04:52:58PM -0700, Trent Piepho wrote: > > On Tuesday, June 30, 2020 6:33:19 PM PDT Drew Fustini wrote: > > > If "pinctrl-single,pins" has 3 arguments (offset, conf, mux), then > > > pcs_parse_one_pinctrl_entry() does an OR operation on conf and mux to > > > get the value to store in the register. > > > > > > > - vals[found].val = pinctrl_spec.args[1]; > > > + > > > + switch (pinctrl_spec.args_count) { > > > + case 2: > > > + vals[found].val = pinctrl_spec.args[1]; > > > + break; > > > + case 3: > > > + vals[found].val = (pinctrl_spec.args[1] | > > pinctrl_spec.args[2]); > > > + break; > > > + } > > > > > > dev_dbg(pcs->dev, "%pOFn index: 0x%x value: 0x%x\n", > > > pinctrl_spec.np, offset, > > pinctrl_spec.args[1]); > > > > If #pinctrl-cells value is greater than 2, nothing will set vals[found].val to > > anything other than zero (from when it's calloc'ed) and the pinctrl will > > silently be programmed to zero. > > If #pinctrl-cells is 3, then it will be: > > vals[found].val = (pinctrl_spec.args[1] | pinctrl_spec.args[2]); > > Do you mean if #pinctrl-cells is great than 3 then it will just have a > default value of zero? > > That does appear to be the case and is probably not the behavior we > want. Thank you for pointing this out. Earlier, there is a check to > make sure there are at least 2 arguments: > > if (pinctrl_spec.args_count < 2) { > dev_err(pcs->dev, "invalid args_count for spec: %i\n", > pinctrl_spec.args_count); > break; > } > > I'll submit a patch where the upper bound is also checked: > > if (pinctrl_spec.args_count < 2 || pinctrl_spec.args_count > 3) { > dev_err(pcs->dev, "invalid args_count for spec: %i\n", > pinctrl_spec.args_count); > break; > } > I was mistaken when I wrote the above. I was using the term #pinctrl-cells when I should have been writing pinctrl_spec.args_count. pinctrl_spec.args_count is 2 when #pictrl-cells is 1. pinctrl_spec.args_count is 3 when #pictrl-cells is 2. I have submitted patches [1][2] with fixes for the bounds check and the dev_dbg(). > > The debug printout was not change to print vals[found].val, so it will > > continue to print the value of the 2nd cell. > > Thank you for pointing this out. Yes, this is an oversight and I will > submit a patch. > > > The result is that a #pinctrl-cells of 3 will produce no warning or error, > > program the pinctrl to zero, whilst at the same time emit debug log messages > > that it is programming the expected values. > > > > The device tree documentation still states that #pinctrl-cells must be 1 when > > using pinctrl-single,pins. This new special case of ORing two values is not > > documented. > > This is a good point, too. I will make a patch to update the > documentation. > > > -Drew [1] https://lore.kernel.org/linux-omap/20200913231557.2063071-1-drew@beagleboard.org/ [2] https://lore.kernel.org/linux-omap/20200913230306.2061645-1-drew@beagleboard.org/
diff --git a/drivers/pinctrl/pinctrl-single.c b/drivers/pinctrl/pinctrl-single.c index f3a8a465d27e..a436ed7762cc 100644 --- a/drivers/pinctrl/pinctrl-single.c +++ b/drivers/pinctrl/pinctrl-single.c @@ -1017,10 +1017,17 @@ static int pcs_parse_one_pinctrl_entry(struct pcs_device *pcs, break; } - /* Index plus one value cell */ offset = pinctrl_spec.args[0]; vals[found].reg = pcs->base + offset; - vals[found].val = pinctrl_spec.args[1]; + + switch (pinctrl_spec.args_count) { + case 2: + vals[found].val = pinctrl_spec.args[1]; + break; + case 3: + vals[found].val = (pinctrl_spec.args[1] | pinctrl_spec.args[2]); + break; + } dev_dbg(pcs->dev, "%pOFn index: 0x%x value: 0x%x\n", pinctrl_spec.np, offset, pinctrl_spec.args[1]);
If "pinctrl-single,pins" has 3 arguments (offset, conf, mux), then pcs_parse_one_pinctrl_entry() does an OR operation on conf and mux to get the value to store in the register. Signed-off-by: Drew Fustini <drew@beagleboard.org> --- drivers/pinctrl/pinctrl-single.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)