Preparing for new release of FC host tools

Mychaela Falconia mychaela.falconia at
Sat Feb 22 21:58:29 UTC 2020

Hi DS,

First of all, thank you for your willingness to test this stuff!

> On Linux 4.19.0-6-amd64 (Debian stable), the problem with the 10ms
> select seems to be not present. With or without the ASYNC_LOW_LATENCY
> patch, the time to dump a flash is the same, eg. 3m45s for the FCDEV3B
> using the pldkit FT2232D at 812500 baud.

Dumping the flash is not the right operation to test for the issue at
hand.  When you execute a loadtool command like dump2bin or flash
dump2bin, loadtool sends a single DUMP command to loadagent, and
loadagent then emits the entire dump (up to 8 MiB) in one fell swoop
without needing further commands from loadtool.  Thus the time it
takes to perform a dump will always be equal to the time it takes
loadagent to emit the full volume of data out through Calypso's UART
transmitter, regardless of how the Linux kernel on the host side
subsequently passes these characters to loadtool, in big or small

Instead the problem which Serg L reported in 2019-07 and which I may
have also seen myself when I tried to run newer Linux-4.4.y versions
(which also had the much bigger problem of causing kernel panics under
non-understood conditions, hence I gave up on them) manifests on
operations like fc-xram loading or flash programming.  When you program
device flash with fc-loadtool (as opposed to dumping it), loadtool
sends one line at a time to loadagent (256 bytes at a time in the case
of flash program-bin or one S-record at a time in the case of
program-m0), then collects the echo from the target and waits for the
loadagent command to complete before sending the next line.  fc-xram
similarly sends each S-record as a separate loadagent command,
collecting the echo and waiting for completion after each line.  For
these operations, if the Linux kernel inserts a delay on select
returning when serial input is ready, the resulting operation time
will be significantly extended.

I would greatly appreciate it if you could test flash programming or
fc-xram loading on your system and see if the ASYNC_LOW_LATENCY change
makes any improvement.  I am otherwise ready to put out the new FC
host tools release, just waiting for these tests.

Hasta la Victoria, Siempre,
Mychaela aka The Mother

More information about the Community mailing list