On Fri, 2012-03-02 at 13:35 +0530, K, Mythri P wrote: > Hi Tomi, > > On Thu, Mar 1, 2012 at 5:01 PM, Tomi Valkeinen <tomi.valkeinen@xxxxxx> wrote: > > Hi, > > > > On Thu, 2012-03-01 at 13:44 +0530, mythripk@xxxxxx wrote: > >> From: Mythri P K <mythripk@xxxxxx> > >> > >> Currently TX_PHY is put to TX_ON(transmission state) on receiving HPD. > >> It just ensures that the TV is connected but does not guarantee > >> that TMDS data lines and clock lines are up and ready for transmission. > >> Which although is very rare scenario has a potential to damage the HDMI > >> port.(A use case where TV is connected to board but it is in different > >> channel would still trigger a HPD but TMDS lines will be down). > > > > When/how does the damaging happen? When the HDMI cable is disconnected > > in the case above? Or does the damage just happen by having the cable > > connected, but the TV on different channel? > > Damage never happens with the cable connected for a long time.. Damage > happens only > when the cable is connected when the PHY is in TX ON state. Which > requires the following sequence to be followed. > 1. Wait for HPD connect > 2. Wait for the PHY connect ( TMDS lines are up) > 3. Enable PHY > We are currently missing step 2. Doesn't HPD already tell us that the cable is connected? > >> + tmds_clk = hdmi_read_reg(hdmi_wp_base(ip_data), > >> + HDMI_WP_WP_DEBUG_DATA); > >> + udelay(15); > > > > This code doesn't make any sense, or the HDMI TRM is wrong. You read the > > same register, HDMI_WP_WP_DEBUG_DATA, four times, assigning the value to > > data0-2/clk. The TRM I have doesn't talk about anything like that. What > > is the code supposed to do? > > > I am not very sure which TRM you have but the HDMI_WP_WP_DEBUG_DATA > can also be used to probe the TMDS lines as well, but i might as well > try pad config and let you know. I have 4460 HDMI TRM vF and 4430 HDMI TRM vL. Both look the same regarding the debug registers. Can you send me the latest ones that explain the DEBUG registers correctly? > > The register HDMI_TXPHY_PAD_CONFIG_CONTROL also has bits for RXDET_LINE. > > Is that something different? > > > >> + } while ((tmds_data0 != tmds_data1 || tmds_data1 != tmds_data2 > >> + || tmds_data1 != tmds_clk) && (loop < 500)); > > > > This is a rather confusing way to do the loop. Why not just use for-loop > > with the timeout, and check the tmds variables inside the loop. Much > > easier to read. > > > > In the worst case the above causes a 7.5ms busy loop in an interrupt > > handler. That's not very good thing. Why is there a need for the loop? > I agree looping in interrupt context is bad.. That was the reason i > had a threaded irq handler and i case even here it is the threaded irq > handler i dont think it is happening in irq context. We should nevertheless avoid 7.5ms busyloop if at all possible. Is there an estimate of how long it takes for the PHY to connect? I think we could just start a delayed work from the interrupt, and do the check after a short delay. I don't think the user cares if the connect happens after 1ms or 10ms from the moment the cable is plugged in =). Tomi
Attachment:
signature.asc
Description: This is a digitally signed message part
