tftp: rootfs.cpio is zero size

from tutorial: HowTo-LCA2018-FPGA-Miniconf-VexRiscv-Renode

I am trying the section “Boot Linux via tftp”. This gist has the full log.

Overview:

$ export CPU=vexriscv CPU_VARIANT=linux PLATFORM=arty TARGET=net FIRMWARE=linux
$ source ./scripts/enter-env.sh
(LX P=arty C=vexriscv.linux F=linux) $
(LX P=arty C=vexriscv.linux F=linux) $ make gateware
(LX P=arty C=vexriscv.linux F=linux) $ ./scripts/build-linux.sh

Here is the part with tftp:

(LX P=arty C=vexriscv.linux F=linux) pdp7@x1:~/dev/litex-buildenv$ make tftpd_stop
# FIXME: This is dangerous...
(LX P=arty C=vexriscv.linux F=linux) pdp7@x1:~/dev/litex-buildenv$ make tftpd_start
mkdir -p build/tftpd/
Starting atftpd
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]: Advanced Trivial FTP server started (0.7)
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   running in daemon mode on port 6069
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   bound to IP address 192.168.100.100 only
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   logging level: 6
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   directory: build/tftpd//
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   user: pdp7.pdp7
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   log file: /dev/stdout
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   not forcing to listen on local interfaces.
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   server timeout: Not used
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   tftp retry timeout: 5
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   maximum number of thread: 100
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   option timeout:   enabled
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   option tzise:     enabled
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   option blksize:   enabled
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:   option multicast: enabled
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:      address range: 239.255.0.0-255
Feb 19 23:35:16 x1 atftpd[16880.139858938175552]:      port range:    1758
(LX P=arty C=vexriscv.linux F=linux) pdp7@x1:~/dev/litex-buildenv$ make tftp
mkdir -p build/arty_net_vexriscv.linux/
time python -u ./make.py --platform=arty --target=net --cpu-type=vexriscv --iprange=192.168.100 -Ob toolchain_path /home/pdp7/dev/litex-buildenv/build/Xilinx/opt/Xilinx/  --cpu-variant=linux --cpu-variant=linux  --no-compile-gateware \
	2>&1 | tee -a /home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux//output.20200219-233522.log; (exit ${PIPESTATUS[0]})
[WARNING] Deprecated, please update : shadow_base replaced by IO regions.
make[1]: warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.
make[1]: Entering directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/libcompiler_rt'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/libcompiler_rt'
make[1]: warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.
make[1]: Entering directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/libbase'
 CC       exception.o
 CC       system.o
 CC       id.o
 CC       uart.o
 CC       time.o
 CC       spiflash.o
 CC       mdio.o
 AR       libbase.a
 AR       libbase-nofloat.a
make[1]: Leaving directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/libbase'
make[1]: warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.
make[1]: Entering directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/libnet'
 CC       microudp.o
 AR       libnet.a
make[1]: Leaving directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/libnet'
make[1]: warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.
make[1]: Entering directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/bios'
 CC       isr.o
 CC       sdram.o
 CC       main.o
 CC       boot.o
 LD       bios.elf
chmod -x bios.elf
 OBJCOPY  bios.bin
chmod -x bios.bin
python -m litex.soc.software.mkmscimg bios.bin --little
make[1]: Leaving directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/bios'
make[1]: warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.
make[1]: Entering directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/uip'
 CC       clock-arch.o
 CC       liteethmac-drv.o
 AR       libuip.a
make[1]: Leaving directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/uip'
make[1]: warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.
make[1]: Entering directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/firmware'
 CC       bist.o
 CC       ci.o
/home/pdp7/dev/litex-buildenv/firmware/ci.c: In function 'status_short_print':
/home/pdp7/dev/litex-buildenv/firmware/ci.c:368:15: warning: unused variable 'underflows' [-Wunused-variable]
  unsigned int underflows;
               ^~~~~~~~~~
/home/pdp7/dev/litex-buildenv/firmware/ci.c: In function 'status_print':
/home/pdp7/dev/litex-buildenv/firmware/ci.c:467:15: warning: unused variable 'underflows' [-Wunused-variable]
  unsigned int underflows;
               ^~~~~~~~~~
At top level:
/home/pdp7/dev/litex-buildenv/firmware/ci.c:1002:21: warning: 'log2' defined but not used [-Wunused-function]
 static unsigned int log2(unsigned int v)
                     ^~~~
 CC       config.o
 CC       encoder.o
 CC       etherbone.o
 CC       ethernet.o
 CC       fx2.o
 CC       hdmi_in0.o
 CC       hdmi_out0.o
 CC       hdmi_out1.o
 CC       heartbeat.o
 CC       isr.o
 CC       main.o
 CC       mdio.o
 CC       mmcm.o
/home/pdp7/dev/litex-buildenv/firmware/mmcm.c: In function 'mmcm_dump_all':
/home/pdp7/dev/litex-buildenv/firmware/mmcm.c:110:6: warning: unused variable 'i' [-Wunused-variable]
  int i;
      ^
 CC       oled.o
 CC       opsis_eeprom.o
bash /home/pdp7/dev/litex-buildenv/firmware/version_data.sh
# Check the version files exist
[ -e /home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/include/../..//software/firmware/version_data.h ]
[ -e /home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/include/../..//software/firmware/version_data.c ]
 CC       pattern.o
 CC       pll.o
 CC       processor.o
/home/pdp7/dev/litex-buildenv/firmware/processor.c: In function 'fb_set_mode':
/home/pdp7/dev/litex-buildenv/firmware/processor.c:532:15: warning: unused variable 'hdmi_out1_enabled' [-Wunused-variable]
  unsigned int hdmi_out1_enabled;
               ^~~~~~~~~~~~~~~~~
/home/pdp7/dev/litex-buildenv/firmware/processor.c:531:15: warning: unused variable 'hdmi_out0_enabled' [-Wunused-variable]
  unsigned int hdmi_out0_enabled;
               ^~~~~~~~~~~~~~~~~
/home/pdp7/dev/litex-buildenv/firmware/processor.c: In function 'processor_service':
/home/pdp7/dev/litex-buildenv/firmware/processor.c:793:29: warning: unused variable 'm' [-Wunused-variable]
  const struct video_timing *m = &video_modes[processor_mode];
                             ^
At top level:
/home/pdp7/dev/litex-buildenv/firmware/processor.c:449:13: warning: 'fb_clkgen_write' defined but not used [-Wunused-function]
 static void fb_clkgen_write(int m, int d)
             ^~~~~~~~~~~~~~~
 CC       reboot.o
 CC       stdio_wrap.o
 CC       tofe_eeprom.o
 CC       uptime.o
 CC       version.o
 CC       pcie.o
 CC       /home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/include/../..//software/firmware/hdmi_in1.o
 LD       firmware.elf
chmod -x firmware.elf
 OBJCOPY  firmware.bin
chmod -x firmware.bin
python -m litex.soc.software.mkmscimg -f --little firmware.bin -o firmware.fbi
make[1]: Leaving directory '/home/pdp7/dev/litex-buildenv/build/arty_net_vexriscv.linux/software/firmware'

real	0m5.673s
user	0m3.277s
sys	0m0.425s
rm -rf build/tftpd/
mkdir -p build/tftpd/
cp build/arty_net_vexriscv.linux//software/linux/firmware.bin build/tftpd//Image
cp build/arty_net_vexriscv.linux//software/linux/riscv32-rootfs.cpio build/tftpd//rootfs.cpio
cp build/arty_net_vexriscv.linux//software/linux/rv32.dtb build/tftpd/
cp build/arty_net_vexriscv.linux//emulator/emulator.bin build/tftpd/

The I run:

make gateware-load
make firmware-connect

meanwhile I get this error:

(LX P=arty C=vexriscv.linux F=linux) pdp7@x1:~/dev/litex-buildenv$ make gateware-load
openocd -f board/digilent_arty.cfg -c "init; pld load 0 build/arty_net_vexriscv.linux//gateware/top.bit; exit"
Open On-Chip Debugger 0.10.0+dev-01012-ged8fa09cf-dirty (2020-01-15-04:14)
Licensed under GNU GPL v2
For bug reports, read
	http://openocd.org/doc/doxygen/bugs.html
Info : auto-selecting first available session transport "jtag". To override use 'transport select <transport>'.
jtagspi_program
Info : ftdi: if you experience problems at higher adapter clocks, try the command "ftdi_tdo_sample_edge falling"
Info : clock speed 10000 kHz
Info : JTAG tap: xc7.tap tap/device found: 0x0362d093 (mfg: 0x049 (Xilinx), part: 0x362d, ver: 0x0)
(LX P=arty C=vexriscv.linux F=linux) pdp7@x1:~/dev/litex-buildenv$ Feb 19 23:37:04 x1 atftpd[16880.139858938087168]: socket may listen on any address, including broadcast
Feb 19 23:37:04 x1 atftpd[16880.139858938087168]: Serving Image to 192.168.100.50:7642
Feb 19 23:37:10 x1 atftpd[16880.139858938087168]: Server thread exiting
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: socket may listen on any address, including broadcast
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: Serving rootfs.cpio to 192.168.100.50:7642
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: Server thread exiting
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: socket may listen on any address, including broadcast
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: Serving boot.bin to 192.168.100.50:7642
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: File build/tftpd//boot.bin not found
Feb 19 23:37:10 x1 atftpd[16880.139858929694464]: Server thread exiting

[mateusz-holenko]

That’s a quite nice analysis of the problem - @pdp7 😃

It looks like some initial wget failed (network problem?) and it created an empty riscv32-rootfs.cpio file. From this moment on the (broken, empty) cached version was served by the tftp server.

The trivial solution would be to not only check if the rootfs file exists, but also verify it’s not empty. I sense more problems though - what if the file is only partially downloaded (i.e., it’s not a proper rootfs image but it’s not empty)?

We should probably add some CRC checking to make sure that files downloaded with wget are correct.

BTW: I also noticed a problem in build-linux.sh script in your log: https://gist.github.com/pdp7/287a0a5159d1877480aaaf6e513c23b1#file-litex-tftp-linux-txt-L3011. It’s not related to your issue, but should be fixed anyway.

<div> Gist</div><div>tftp linux litex</div><div>tftp linux litex. GitHub Gist: instantly share code, notes, and snippets.</div>

[mateusz-holenko]

#347 added checksum verification of downloaded files, so this problem should not happen again.