Micka, Yes, you're right, the network can still work after "Phy not found" was shown, although I observed this situation very rarely (I guess, only once), when I simply powered the board up. But this situation can be created artificially, if you intentionally change PHY address to a non-zero value, as I described in my previous message (there is probably some bug in U-Boot). It's just convenient to call this problem "phy not found".
As for the PRCM.PRM_RSTTIME, I'm studying the U-Boot code and so far don't find any place where this register is written, except for some code apparently intended to be run on some other processors, not on am33xx. And I think the macro CONFIG_OMAP_PLATFORM_RESET_TIME_MAX_USEC is also related to some other boards or processors, because the code where it is used assumes that the PRM_RSTTIME register contains some number of 32.768 kHz clock cycles, in bits [9:0] RSTTIME1, while the same register on AM335x contains some number of 24 MHz clock cycles (CLK_M_OSC) in bits [7:0] RSTTIME1. I need to read the datasheet more to understand better what it is. Regards Alex On Wednesday, December 10, 2014 12:13:46 PM UTC+1, Mickae1 wrote: > > From my experience: "Phy not found" doesn't mean that your network is not > working ... I don't know what Robert Nelson is doing at boot, but it's > possible that he fix this problem by using the register PRCM.PRM_RSTTIME= > 0xFFFF ? > > Micka, > > > Le Tue Dec 09 2014 at 21:40:11, <[email protected] <javascript:>> a > écrit : > >> KeOu, >> >> Thank you for the data. It looks like "phy not found" does not happen >> very often with your boards. >> >> I also run that "update_kernel.sh" to update the kernel to >> 3.8.13-bone68. And I tried 3.14.22-ti-r31 too, and had a lot of "phy not >> found" in this case as well. In all my experiments the console port was >> always connected to a PC. >> >> Regards >> >> Alex >> >> >> On Tuesday, December 9, 2014 6:10:37 PM UTC+1, KeOu Chao wrote: >>> >>> Alex, >>> >>> I did more testing by using a remote power switch; using 3 BBBs with 3 >>> different Kernel, no modification on U-Boot, Kernel. >>> >>> - 3.8.13-bone47 - that came with the board >>> - 3.8.13-bone68 - upgrade via on board script /opt/scripts/tools/update_ >>> kerenl.sh >>> - 3.14.22-ti-r31 - download from run-ee.net >>> >>> My test scripts turn power switch wait for 50 seconds, send TCP SYN >>> request per second for 5 times, if no ACK received, declare Failure; then >>> power off for 5 seconds, repeat. >>> >>> One thing I found is that with Console port connected, there was no >>> failure on all my tests (other then bone47). Do you have console port >>> connect when testing ? >>> >>> This is my test results, each test has 600 power cycles, >>> >>> Test 1 Test 2 Test 3 Test 4 >>> 3.8.13-bone47 50 8.33% (a) 46 7.76% (a) 41 6.83% (a) 40 6.67% (b) >>> 3.8.13-bone68 20 3.33% (a) 0 0% (b) 0 0% (b) 21 3.5% (a) >>> 3.14.22-ti-r31 0 0% (b) 27 4.5% (a) 0 0% (c) 24 4% (a) >>> >>> (a) no console connected >>> (b) with console connected >>> (c) with patch - DGGND to console GND, VDD_3V3 to Console Tx, VDD_3V3 to >>> Console Rx. >>> >>> I just start looking into this, and have not digest all the notes / >>> manual yet; Hope the information help. >>> >>> Regards >>> >>> KeOu >>> >>> >>> Hi KeOu, >>> >>> I guess with 3.8.13-bone67 I had "phy not found" as frequently as with >>> the original image. On one board with 3.8.13-bone67, I had just one >>> occurrence of "phy not found" in 50 power-on resets. On the other board >>> with the same kernel I had 22 "phy not found" out of 50 power-on resets. >>> Then I updated the kernel to 3.8.13-bone68, and still had that error, 12 >>> times out of 50 power-on resets. >>> >>> To perform that kernel update, I followed the instructions from the >>> error message that showed up after I attempted to run "install-me.sh" from >>> http://rcn-ee.net/deb/wheezy-armhf/v3.8.13-bone68/ Could you please >>> describe in more detail what prebuilt image you tried and how exactly you >>> installed that? Did you also install some other U-Boot? >>> >>> Some time ago, I installed and tried one of the rcn-ee.net images >>> following a procedure similar to that described here >>> <http://www.armhf.com/boards/beaglebone-black/bbb-sd-install/>. But it >>> didn't help. >>> >>> There is one important thing to remember: "phy not found" happens before >>> autoboot in U-Boot, and if that happens, no matter which Linux is loaded >>> afterwards - the transceiver chip doesn't work properly. So, both >>> 3.8.13-bone67 and 3.8.13-bone68 fail to detect the chip, if it wasn't >>> detected by U-Boot. >>> >>> Regards, >>> Alex >>> >>> On Saturday, December 6, 2014 5:59:56 PM UTC+1, [email protected] wrote: >>>> >>>> Hi Alex, >>>> >>>> In past couple days, I went through the same path as you did. However, >>>> with my limited tests, I see improvement on 3.8.13-bone68. I have BBB Rev >>>> C >>>> from Element14 with a generic 5V 2.8A switch power supplier. >>>> >>>> - with 3.8.13-bone47 (the original image) power cycle 50 times, there >>>> were 2 "phy not found" >>>> - with 3.8.13-bone68 (prebuild image from http://rcn-ee.net/, power >>>> cycle 50 times, there were no "phy not found" >>>> I also tested 3.14.22-ti-r31, power cycle 50 time, and no "phy not >>>> found". >>>> >>>> I'll do more test to see if the issue show up. In you tests, did you >>>> see any improvement in 3.8.13-bone67 vs earlier version ? >>>> >>>> Thanks >>>> >>>> KeOu >>> >>> >>> -- >> For more options, visit http://beagleboard.org/discuss >> --- >> You received this message because you are subscribed to the Google Groups >> "BeagleBoard" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected] <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > -- For more options, visit http://beagleboard.org/discuss --- You received this message because you are subscribed to the Google Groups "BeagleBoard" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
