[PATCH v2 1/2] commands: change Y-Modem implementation

[Robert Jarzmik]

Jean-Christophe PLAGNIOL-VILLARD <plagnioj at jcrosoft.com> writes:
> On 19:07 Mon 05 Nov     , Robert Jarzmik wrote:
>> Jean-Christophe PLAGNIOL-VILLARD <plagnioj at jcrosoft.com> writes:
>> No timeout indeed, on purpose.
>> 
>> The short explanation is that in a purge situation, all incoming data must be
>> flushed, regardless of the time it takes.
>> 
>> The long explantion is :
>>  - suppose the while(cdev->tstc(cdev)) takes a lot of time
>>  - this implies that :
>>    => the sender is sending data as fast as the reader can consume (not quite
>>    probable, but why not ...)
>>    => the sender has gone crazy, as in *-Modem protocol when this function is
>>    called, no more than 1029 bytes can be sent by a *sane* sender
>> 
>> So let's take as granted that the sender has gone crazy (or hardware is brain
>> dead). What will happen if I place a timeout here ?
>>  - first, I'll obviously get out of xy_flush()
>>  - then, as I receive garbage, xy_handle() will exit with either -EBADMSG or
>>  -EILSEQ. So far so good.
>>  - then the loady command will finish on error. Still good.
>>  - then the input console will take over, and execute all the garbage sent to
>>  it.
>>    This is definitely something we don't want. Imagine in the garbage you have
>>    something like "erase /dev/mtd0" ... Therefore, no timeout.
> my issue is how can I interrupt it from barebox
You can't, how could you ? Your console is overwhelmed by garbage input, what
could you possibly do to interrupt it apart from cutting down the sender ?

> can you put the v3 on a tree somewhere so I can pull it
I'll try. I have no external git I can push to. If you have a hint ...

-- 
Robert