2009-07-13 15:44:34 read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Amol Sukerkar (UNITED STATES)
Message: 77145
Hello,
I am developing an application that executes a process using execve call and talks to it using Inter Process Communication pipes. I have noticed that the read and write system calls respond slowly and take as much as 1 - 2 seconds to return on 0.4 and 0.3 silicon versions of BF533. It works just fine without any delay with silicon version 0.5. This looks as a hardware issue. Is this a known issue? If it is, then, is there a workaround associated with it? Any help would be appreciated.
Information
Platform: BF533 (0.3, 0.4, 0.5)
Filesystem: Blackfin uClinux (Kernel Version: 2.4.8 antique?)
QuoteReplyEditDelete
2009-07-13 16:23:39 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Robin Getz (UNITED STATES)
Message: 77146
Amol:
Update your kernel. Are you really using 2.4?
-Robin
QuoteReplyEditDelete
2009-07-13 19:49:44 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Amol Sukerkar (UNITED STATES)
Message: 77149
Thanks, Robin! I wish it was in my capacity to update the kernel (I would have done it years ago). Anyways, the reason I was baffled about this was because the read, write work fine with 0.5 silicon revision and not with 0.3 or 0.4 silicon revisions. So, could this be a hardware issue and not kernel issue? I was wondering if the uClinux maintainers were aware of this issue and had any pointers for me to look in the right direction.
- Amol
QuoteReplyEditDelete
2009-07-14 00:09:51 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Mike Frysinger (UNITED STATES)
Message: 77153
linux-2.4 has never come from us. we have no idea what's going on with it.
QuoteReplyEditDelete
2009-07-14 08:14:28 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Robin Getz (UNITED STATES)
Message: 77213
Amol:
It is likely that you are running into an anomaly that existing on 0.3 and 0.4, but not on 0.5 - consider yourself luckly that things work at all.
Please let me know what product this is in - so I can know not to buy it. the 2.4 kernel is not safe to use.
-Robin
QuoteReplyEditDelete
2009-07-14 11:25:36 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Amol Sukerkar (UNITED STATES)
Message: 77235
That was very helpful. I am quiet sure it is a silicon anomaly and nothing to do with linux kernel, hence, I decided to post on the forum. Apart from this, I don't have any comments. I'll go to the gym instead.
QuoteReplyEditDelete
2009-07-14 11:30:08 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Mike Frysinger (UNITED STATES)
Message: 77238
you're asking about a code base that has never come from us, that we've never seen, and that we have no idea about. basically, with linux-2.4, you're completely on your own. or go bug the people who gave you the source code in the first place.
QuoteReplyEditDelete
2009-07-14 12:33:03 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Amol Sukerkar (UNITED STATES)
Message: 77248
Mike,
This is a misunderstanding. I am not talking about source code at all. I've been posting about my crappy 2.4 code for quiet some time now and get the same answer. Upgrade to 2.6. I posted this to find out if this hardware anomaly is known to the maintainers of uClinux.
QuoteReplyEditDelete
2009-07-14 12:43:38 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Mike Frysinger (UNITED STATES)
Message: 77251
there is no way to even guess at what is going wrong in your system. read()/write() functions do not directly touch the hardware in anyway.
you keep getting the same answer because 2.4 is different from 2.6 in pretty much everything
QuoteReplyEditDelete
2009-07-14 17:33:37 Re: read, write system calls slow with BF533 0.4, 0.3 silicon rev.
Robin Getz (UNITED STATES)
Message: 77263
Amol:
So - why would you expect a different answer?
As far as I know -- there were zero anomaly workarounds included in the 2.4 kernel. If you want to look out for things - check out the anomaly list:
https://docs.blackfin.uclinux.org/doku.php?id=bfin:anomalies
Your kernel will have zero workarounds - unless you added them.
It also just plain did not work well.
-Robin