RetroBSD
http://retrobsd.org/

LiteBSD - panic microtime watchdog
http://retrobsd.org/viewtopic.php?f=16&t=37323
Page 1 of 1

Author:  Pito [ Thu Oct 15, 2015 8:26 pm ]
Post subject:  LiteBSD - panic microtime watchdog

Code:
login: root
Copyright (c) 1980, 1983, 1986, 1988, 1990, 1991, 1993, 1994
        The Regents of the University of California.   All rights reserved.

4.4BSD-Lite UNIX #1: Fri Apr 01 00:00:00 PDT 1994

Welcome to 4.4BSD-Lite!

panic: microtime watchdog
syncing disks... done

dumping to dev 2, offset 64512
dump device bad

Maybe an sdcard problem - an older Kingston 4GB

Author:  Pito [ Thu Oct 15, 2015 8:31 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

With the oldest 512MB Kingston:
Code:
4.4BSD-Lite (bsd.net) (console)

login: root
Copyright (c) 1980, 1983, 1986, 1988, 1990, 1991, 1993, 1994
        The Regents of the University of California.   All rights reserved.

4.4BSD-Lite UNIX #1: Fri Apr 01 00:00:00 PDT 1994

Welcome to 4.4BSD-Lite!

erase ^H, kill ^U, intr ^C status ^T
Don't login as root, use the su command.
# cd /
# ls -l
total 48
-rw-rw-r--  1 root  wheel   684 Oct 10 12:48 .cshrc
-rw-rw-r--  1 root  wheel   286 Oct 10 12:48 .profile
-rw-rw-r--  1 root  wheel  3425 Oct 10 12:48 COPYRIGHT
drwxrwxr-x  2 root  wheel   512 Oct 10 12:48 bin
drwxrwxr-x  3 root  wheel  2048 Oct 10 12:48 dev
drwxrwxr-x  3 root  wheel  1024 Oct 10 12:48 etc
drwxrwxr-x  2 root  wheel   512 Oct 10 12:48 home
drwxrwxr-x  2 root  wheel   512 Oct 10 12:48 root
drwxrwxr-x  2 root  wheel  1024 Oct 10 12:48 sbin
drwxrwxrwt  2 root  wheel   512 Oct 10 12:48 tmp
drwxrwxr-x  9 root  wheel   512 Oct 10 12:48 usr
drwxrwxr-x  7 root  wheel   512 Oct 10 12:48 var
#

Author:  vak [ Thu Oct 15, 2015 8:34 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Pito wrote:
Maybe an sdcard problem - an older Kingston 4GB
This happens, when some driver disabled interrupts for too long, more than a second.

Probably your SD card is too slow to respond or something.
From my experience, Samsung microSDHC 8gb cards (Class 6 or EVO) work the best for LiteBSD.

Author:  Pito [ Thu Oct 15, 2015 8:39 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Tomorrow I am going to buy
SanDisk Micro SDHC ultra card 8GB 48MB/s CL10
Samsung Micro SDHC EVO 8GB Class 10 UHS-I
- they may work with high speed CMD6 as well.

Author:  vak [ Thu Oct 15, 2015 8:42 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

It will be good to test these cards using diskspeed utility and compare with the table on page: http://retrobsd.org/wiki/doku.php/doc/sd-benchmark.

Author:  Pito [ Thu Oct 15, 2015 8:50 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

My current and fastest 512MB Kingston @12MHz SPI
Code:
/var/spool/uucp:
total 0
-r--r--r--  1 root  wheel  0 Oct 10 12:48 ERRORS

/var/tmp:
       49.53 real         8.73 user         0.82 sys
# diskspeed
Testing 4-kbyte block size.
Write speed: 8 Mbytes in 8.854 seconds = 925 kbytes/sec
 Read speed: 8 Mbytes in 6.052 seconds = 1353 kbytes/sec
#

Author:  vak [ Thu Oct 15, 2015 8:52 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Not bad.
How about "time ls -lR /"?

Author:  Pito [ Thu Oct 15, 2015 8:56 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

See above (before diskspeed). Mind it is @12MHz and your benchmarks are @16.5MHz.

Author:  vak [ Thu Oct 15, 2015 8:58 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

This SD benchmark table had been created for 16MHz SD clock.
You have a pessimistic 12MHz setup in the kernel, though.
In next LiteBSD version, I will increase the SD clock for SDZL and WiFire boards, so it will be even faster for you.

Author:  vak [ Thu Oct 15, 2015 9:00 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Pito wrote:
My current and fastest 512MB Kingston @12MHz SPI
Code:
       49.53 real         8.73 user         0.82 sys

This is too much.
Good values should be somewhere in range 16-17 seconds.

Author:  Pito [ Thu Oct 15, 2015 9:00 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

My first impression from quick messing with it is - it is as fast as retrobsd.. :)
sl even slower.. Why the "sl" does not move fluently?

Author:  vak [ Thu Oct 15, 2015 9:07 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Pito wrote:
My first impression from quick messing with it is - it is as fast as retrobsd.. :)
sl even slower.. Why the "sl" does not move fluently?

It's a demand paging.
The program's code is being moved into memory not all at once, like in RetroBSD, but gradually, as the execution progresses.

Author:  Pito [ Thu Oct 15, 2015 9:19 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

This is something which needs to be explained to me.
Long time back I indicated (and stopped working with LiteBSD therefore) that I am not able to run the stuff similarly as on the retrobsd.
For example I was not able to run several jobs in background.
So how it actually works?

Author:  vak [ Fri Oct 16, 2015 3:51 am ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Pito wrote:
This is something which needs to be explained to me.
Long time back I indicated (and stopped working with LiteBSD therefore) that I am not able to run the stuff similarly as on the retrobsd.
For example I was not able to run several jobs in background.
So how it actually works?

Well, it does work, just don't expect the system to be as stable as RetroBSD yet.

Demand paging means that when the process starts, the program code is not required to be present in memory. The MMU block (memory management unit, sometimes called TLB) available in MZ processor, divides the memory into pages, and allows to individually map physical pages into virtual address space of user processes. At start, only one page is allocated for the process: a top of stack, which also contains the command line arguments passed to it from the parent process. The code segment is mapped to a binary file. On the vey first instruction of the program, a so called TLB fetch exception happens, and the kernel allocates a physical page, reads the code page (4 kbytes) from the binary file and maps the page to the virtual address of the process. Then the execution continues.

User pages are cached in kernel i/o buffers, so for repeating processes there is no SD card i/o at all. No need to swap out code pages, as they are read-only and can be always read in from the binary file when requred. This is a big advantage of LiteBSD over FreeBSD, where the shell mush be swapped out for every next process, which causes a lot of i/o and may result in SD card wear out.

For a good explanation of how MMU is organized I would recommend Chapter 6 "Low-level Memory Management and the TLB" in the book "See MIPS Run", second edition.

Even more detailed information is available in M5150 Processor Core Family Software User’s Manual, Chapter 6 "Memory Management of the M5150 Core". You can get this PDF from Imagination website (with registration) via link: http://imgtec.com/?do-download=4174.

Author:  Pito [ Fri Oct 16, 2015 6:10 am ]
Post subject:  Re: LiteBSD - panic microtime watchdog

So would "sl" work fluently when LiteBSD will be stable enough?

Author:  vak [ Sun Oct 18, 2015 3:02 am ]
Post subject:  Re: LiteBSD - panic microtime watchdog

Pito wrote:
So would "sl" work fluently when LiteBSD will be stable enough?

"sl" is fine by itself.
It's the FT232R dongle who causes this effect of uneven output.
With dongles of other types like CH340G, or PL-2303HX, or CP2102 it works pretty smoothly.

Author:  Pito [ Sun Oct 18, 2015 6:49 am ]
Post subject:  Re: LiteBSD - panic microtime watchdog

It works fine with PL2303TA. I thought FT232r is a gold standard :)

Author:  vak [ Sun Oct 18, 2015 7:02 pm ]
Post subject:  Re: LiteBSD - panic microtime watchdog

It was a surprise for me as well. From the driver's quality point of view, seems like CP2102 is the best alternative. PL-2303 driver causes a crash on my iMac whenever I plug out the dongle with pending kermit connection.

Page 1 of 1 All times are UTC
Powered by phpBB® Forum Software © phpBB Group
https://www.phpbb.com/