sfx2000

Members
  • Content Count

    559
  • Joined

  • Last visited


Reputation Activity

  1. Like
    sfx2000 reacted to Igor in Armbian v20.05 (Kagu) Planning Thread   
    For those who couldn't made it, meeting logs: https://freenode.irclog.whitequark.org/armbian/2020-04-04 Meeting summary goes directly into Jira issues/bugs and the actual work. 
     
    Thank you all for attending the meeting!
  2. Like
    sfx2000 got a reaction from 5kft in Orange Pi Zero Plus2 H5 hardware oddity in VDD_CPUX power circuit   
    Recently revisiting things with H5 stability under certain SMP load factors* with @5kft and assistance with @lanefu - and H5 tends to be stable around 1104Mhz CPU/504MHz DDR with a 1.3v regulator overlay, and 1008/504 at 1.1v on NanoPi NEO2 - which doesn't really use the Mali450 at all.
     
    @lanefu was using a different H5 board (Orange Pi something or other - pls comment) - but he was also able to reproduce the stability issue I noted.
     
    Might be more interesting on something that engages all cores on the CPU, and does the GPU stress as well.
     
     
    * openssl speed -multi 4 - which flexs the ARMv8 enhancements there, engaging more of the logical blocks in Cortex-a53 - note that the governor in use is schedutil, which tends to be a different code path than "performance" or "on demand"
     
    @5kft - has been super awesome at helping out - reproducing the issue, concurring with analysis, proposing fixes...
     
    sfx
  3. Like
    sfx2000 reacted to martinayotte in How enable SPI slave mode?   
    Most SoC are not "slave capable" but only "master" ...
  4. Like
    sfx2000 reacted to Da Xue in Orange Pi Zero Plus2 H5 hardware oddity in VDD_CPUX power circuit   
    @tkaiser Allwinner only certifies the H3 to operate at 1008MHz @1.2V and H5 to operate at 1008MHz @ 1.1V with DDR clocks up to 672MHz. Designs and software support for adjustable voltage and higher clock rates must be validated by the third party that chooses to implementing such features.  DVFS will not be supported in the Allwinner H3/H5 BSPs since the transition to AXP8036. I have seen some small boards with 16-bit DDR3 (1 DDR chip) clocked at 744MHz but they are almost guaranteed to have memory errors and video playback issues.
  5. Like
    sfx2000 got a reaction from lanefu in PREEMPT-RT PATCH for Allwinner H5   
    Curious as to why one wants to apply the patches for preempt-rt in the first place? Do you have a reason/application that would benefit? Most times, one can see an overall decrease in application performance, and to take advantage of preempt-rt, the application must call pthreads in a very specific way - which means that many times, the app must also be rebuilt.
     
    My reason for asking - if preempt-rt was so awesome, it would already be in the upstream linux kernel mainline and not require the patch. Avoiding politics around preempt-rt - many of the benefits here can be done without the patches, by selecting an appropriate scheduler - schedutil looks fairly interesting at the moment.
     
    I work on networking oriented applications that have hard deadlines, on MIPS, PowerPC, ARM (both 32bit ARMv7a and 64-bit), and never had a need to apply the preempt-rt patches.
     
    By networking oriented applications - I'm saying 20mSec framing for SIP-RTP (VoIP apps), along with 3GPP/UMTS signalling - if that ain't close to real-time, I don't know what is.
  6. Like
    sfx2000 got a reaction from lanefu in Placemaker - H5 crashing under SMP load   
    In any event, the clock diffs from "stable" to "unstable" - performance overall isn't enough to justify the risks, unless one looks at specific benchmarks...
     
    I'm not into benchmarking - I've got an interest in operational usage of the device.
     
    just my thoughts...
     
    sfx
  7. Like
    sfx2000 reacted to lanefu in Placemaker - H5 crashing under SMP load   
    Well.... that seemed to take it down pretty quick

     
  8. Like
    sfx2000 got a reaction from 5kft in Placemaker - H5 crashing under SMP load   
    Sounds good - and folks should test around the 504MHz DDR clock, along with the overlay to upclock on boards that support the 1.3V regulator... both at 1.2 and 1.3 GHz.
  9. Like
    sfx2000 reacted to 5kft in Placemaker - H5 crashing under SMP load   
    OK well that answers that...  I think it's clear that the memory tests used back in 2017 weren't sufficient to determine the stability of this clock.  Why don't I go ahead and set it to 504MHz as that's the FA default, then if desired people could look at overclocking this further.
  10. Like
    sfx2000 got a reaction from 5kft in Placemaker - H5 crashing under SMP load   
    ok - so the stock FA image also crashes on the same test - it behaves differently than the Armbian image, as it kills off the threads when it tries to do a privileged memory access...
     
    since we're working with armv8-a, we have kernel space (EL1) and user space (EL0) - hence the data abort, as the memory is marked as EL1, and an EL0 task cannot access that. I think that overclocking the CPU exposes a bug that is latent, even without the overlay, and this goes not to device tree, but to uboot and DDR ram init vectors there.
     
    The stress test (openssl) can show the bug, but this isn't the real problem, and the overlay just enables it to happen faster - getting board temp to around 60c, which on a small board like this, includes not only the SoC, but the DDR, can accelerate this issue, as some DDR can get a bit unstable at that temp.
     
    I don't have much time right now to debug further, as I'm in the middle of sfx's North America Tour - last week Austin, TX, next week Miami, FL, Atlanta, GA, Denver, CO, and a short trip to Salt Lake City, UT - about a week of downtime in the SAN, then back to Austin for a week.
     
    @5kft  -- Gnarly problem to sort, eh? But spending time might help other AW H5 targets...
     
    @Igor -- something to watch maybe
  11. Like
    sfx2000 got a reaction from Igor in Placemaker - H5 crashing under SMP load   
    ok - so the stock FA image also crashes on the same test - it behaves differently than the Armbian image, as it kills off the threads when it tries to do a privileged memory access...
     
    since we're working with armv8-a, we have kernel space (EL1) and user space (EL0) - hence the data abort, as the memory is marked as EL1, and an EL0 task cannot access that. I think that overclocking the CPU exposes a bug that is latent, even without the overlay, and this goes not to device tree, but to uboot and DDR ram init vectors there.
     
    The stress test (openssl) can show the bug, but this isn't the real problem, and the overlay just enables it to happen faster - getting board temp to around 60c, which on a small board like this, includes not only the SoC, but the DDR, can accelerate this issue, as some DDR can get a bit unstable at that temp.
     
    I don't have much time right now to debug further, as I'm in the middle of sfx's North America Tour - last week Austin, TX, next week Miami, FL, Atlanta, GA, Denver, CO, and a short trip to Salt Lake City, UT - about a week of downtime in the SAN, then back to Austin for a week.
     
    @5kft  -- Gnarly problem to sort, eh? But spending time might help other AW H5 targets...
     
    @Igor -- something to watch maybe
  12. Like
    sfx2000 reacted to hexdump in Choice of TV box.   
    this is just a quick note that in my experience the amount of tv boxes with fake specs has grown quite a bit in the last months and that this is something to always have in mind when getting a box for a surprisingly cheap (i.e. quite a bit cheaper than usual or most of the other offerings) price - you might be lucky and it will be a bargain or you might hit one with fake specs. some examples i saw recently: a qplus 4g ram / 32g emmc ended up to be 2g ram and 16g nand, a h6 box sold as 4g ram / 32g emmc ended up as 2g ram / 16g emmc, a x96mini 2g ram / 16g emmc ended up at only 1g ram / 16g emmc, a r39 2g ram / 16g emmc with rockchip rk3229 ends up as 1g ram / 16g emmc and an allwinner h3 cpu and so on. the fake specs are not that easy to spot: in android they even fake the storage size shown in the storage settings and with a terminal installed even the "free" command tells you most of the time that the memory amount is proper. what usualy works for storage is "cat /proc/partitions" and watching for the device itself (for instance mmcblk0) - this also quickly shows you if its emmc (=mmcblk) or nand (=nand) and for memory "dmesg | grep -i mem" (do this immediately after booting android, otherwise the memory lines from the bootup might run out of the log buffer) - both of course called in a terminal app. booting one of balbes150's armbian images usually quickly shows you the real specs of the box too.
     
    good luck at not ending up with fake boxes and best wishes - hexdump
     
    p.s.: one thing to keep in mind is that allwinner h6 boxes always only can use 3g ram, even if they have 4g installed - this is a limitataion of the soc ...
  13. Like
    sfx2000 got a reaction from TRS-80 in Surveying the hardware landscape 2019 and beyond, with an eye toward freedom (headless server)   
    Without getting into the file system wars - might consider BTRFS, which is much more GPL friendly...
     
    One of the challenges with the low memory platforms is that ZFS is fairly RAM and compute intense for what it does - that being said, for the right purpose, it's a good file system to look at.
  14. Like
    sfx2000 got a reaction from TRS-80 in Surveying the hardware landscape 2019 and beyond, with an eye toward freedom (headless server)   
    One of the side benefits of China becoming a surveillance society - Government needs lots of cameras to keep an eye on the folks over in Xinjiang....
  15. Like
    sfx2000 got a reaction from Jack953 in free software supported wifi card phone usable ESP8089 ESP8266 ESP32?   
    It's really hard to do in a power efficient manner - there are SDR's that can do all the right waveforms, the signalling, etc, but most of this is on an FPGA - the UMTS and LTE protocols are fairly complex, much more so that 802.11 wifi, and we know how hard that can be.
     
    It's also a massive minefield of patents, which makes GPL are very real challenge.
     
    Osmocom.org has made the most progress towards a free modem - https://osmocom.org/
     
     
  16. Like
    sfx2000 got a reaction from NicoD in Poll : What board images do you trust to use?   
    Depends on the vendor - some do a better job than others...
     
    And there, depends on the SoC's being used.
     
    Not naming names there to avoid slanting the poll...
  17. Like
    sfx2000 got a reaction from Ali-leb in M1+ wifi   
    Broadcom BCM43438 Wi-Fi 802.11n (2.4GHz only) + Bluetooth 4.1 (Dual Mode) combo chip
  18. Like
    sfx2000 got a reaction from suberimakuri in RK3328 Kernel   
    Hmm... actually yes - is the role of Armbian to teach users about Linux? Probably not, IMHO, they're better served by the Pi Folks - Raspbian is good training wheels for folks dipping their toes into Linux
     
     
    The average Armbian user is not a paying customer - there are no Service Level Agreements dictating that any bug of a defined severity level must be fixed within a specific timeframe.
     
    So yes - just ignore it for the most part, step in if is sounds fairly interesting - sounds a bit mean spirited perhaps, but generally folks will step in to help out, and if it's really a bug that is directly traceable back to Armbian code - then if it pops up enough, fix it directly, or delegate it to someone who is maintaining it (better to delegate, as that person knows the code likely better).
     
    So if someone has problematic hardware - "my whoflungpI Zero2W with the built in XYZ WiFi adapter doesn't work" - well, it might be crap hardware, and in the sub-$50USD field of hacker boards and TV boxes, there's a fair amount of crap - can't fix bad hardware, and there, someone will tell them, use something that does work.
  19. Like
    sfx2000 got a reaction from guidol in [Info] colored bash-prompt   
    And for those who live on the command line - byobu is like screen and tmux on steroids - and it's in the ubuntu repos...
     
     
    If one wants to have a fun demo...
     
     
    byobu rocks...
  20. Like
    sfx2000 got a reaction from Igor in RK3328 Kernel   
    Hmm... actually yes - is the role of Armbian to teach users about Linux? Probably not, IMHO, they're better served by the Pi Folks - Raspbian is good training wheels for folks dipping their toes into Linux
     
     
    The average Armbian user is not a paying customer - there are no Service Level Agreements dictating that any bug of a defined severity level must be fixed within a specific timeframe.
     
    So yes - just ignore it for the most part, step in if is sounds fairly interesting - sounds a bit mean spirited perhaps, but generally folks will step in to help out, and if it's really a bug that is directly traceable back to Armbian code - then if it pops up enough, fix it directly, or delegate it to someone who is maintaining it (better to delegate, as that person knows the code likely better).
     
    So if someone has problematic hardware - "my whoflungpI Zero2W with the built in XYZ WiFi adapter doesn't work" - well, it might be crap hardware, and in the sub-$50USD field of hacker boards and TV boxes, there's a fair amount of crap - can't fix bad hardware, and there, someone will tell them, use something that does work.
  21. Like
    sfx2000 got a reaction from Igor in RK3328 Kernel   
    Always hard to mask HW w/SW, but that's always been true - and sometimes it's upstream at a chip level errata.
     
    Don't sell the community here short - there are plenty of active contributors here that have specific skills, and some do this kind of work in their day jobs - but it is a valid point that it's unpaid, volunteer time, and resources are never enough.
     
    The fact that Armbian is on the radar as a serious distro is a good sign of the quality of the community.
  22. Like
    sfx2000 got a reaction from gprovost in Helios64 Annoucement   
    16MB SPI NOR is a good choice, not just for uBoot, but one could put an entire operating system in there (openwrt for example)
     
     
    I agree - @chwe - solder down is going to offer benefit of cost and board space - 16GB is plenty of space considering the other connectivity.
     
    @gprovost - nice board...
  23. Like
    sfx2000 got a reaction from markbirss in Very Small Platforms - Rockchip 3308 and Allwinner V3s   
    Have been exploring Dingleberry Pi 2
     
    QC/Atheros IPQ4019 - more relevant here for the ARM folks - IPQ4019 is a quad core Cortex-A7@800MHz
     
    Buys us GigE on WAN/LAN ports, along with ATH10K for dual-band 802.11 a/b/g/n/AC Wave2 along with USB3.0 for the USB user facing port.
     
    BOM cost is higher, as it manf. costs with more layers, and more power needs - and numbers there do not make sense for an affordable board for this community - $100 GOGS there...
  24. Like
    sfx2000 got a reaction from markbirss in Very Small Platforms - Rockchip 3308 and Allwinner V3s   
    Quick update on DingleBerry Pi...
     
    Schematics - done - It's essentially a respin of the QC/Atheros reference design into a Pi board form factor.
    Layout/Gerbers - done - 4 layer board, single sided - nice to have a friend with an Orcad license from Cadence.
     
    Specifics
    MIPS24kc @ 650Mhz - big endian 802.11n - ATH9K driver - 2*2:2 for 300Mbps at 2.4GHz (wide channels) - PCB trace antennas Two 100Base-2 Ethernet - WAN dedicated port, LAN on switched port One USB-A for peripherals MicroUSB for Power/Console 32MB SPI-NOR flash - uboot and OS 128MB SPI-NAND flash - extending FS for application and user space 64MB DDR2 RAM SW - OpenWRT Master on the ATH79 target Bootloader - uboot with Pepe2K mods (web server fail-safe)  
    The 4351 does have PCIe, but I've decided not to implement
     
    Performance Targets:
    Reference board is good for 100Mbps WiFi and NAT LAN/WAN performance so fairly balanced there. Power - 5V5DC, 1000ma for PS - right now with everything active on reference design, we're around 850ma at max load.  
    BOM is reasonable, COGS says we're around $50USD to cover the NRE and breakeven with 1K boards with kits (box, cables, boards, ac adapter)
     
    I will not build/ship at a loss... So the next step before doing the engineering sample boards it to gauge the potential market interest.
  25. Like
    sfx2000 got a reaction from TonyMac32 in Very Small Platforms - Rockchip 3308 and Allwinner V3s   
    Quick update on DingleBerry Pi...
     
    Schematics - done - It's essentially a respin of the QC/Atheros reference design into a Pi board form factor.
    Layout/Gerbers - done - 4 layer board, single sided - nice to have a friend with an Orcad license from Cadence.
     
    Specifics
    MIPS24kc @ 650Mhz - big endian 802.11n - ATH9K driver - 2*2:2 for 300Mbps at 2.4GHz (wide channels) - PCB trace antennas Two 100Base-2 Ethernet - WAN dedicated port, LAN on switched port One USB-A for peripherals MicroUSB for Power/Console 32MB SPI-NOR flash - uboot and OS 128MB SPI-NAND flash - extending FS for application and user space 64MB DDR2 RAM SW - OpenWRT Master on the ATH79 target Bootloader - uboot with Pepe2K mods (web server fail-safe)  
    The 4351 does have PCIe, but I've decided not to implement
     
    Performance Targets:
    Reference board is good for 100Mbps WiFi and NAT LAN/WAN performance so fairly balanced there. Power - 5V5DC, 1000ma for PS - right now with everything active on reference design, we're around 850ma at max load.  
    BOM is reasonable, COGS says we're around $50USD to cover the NRE and breakeven with 1K boards with kits (box, cables, boards, ac adapter)
     
    I will not build/ship at a loss... So the next step before doing the engineering sample boards it to gauge the potential market interest.