Writing to FLASH memory takes application's CPU

[thehadar]

Hi guys,

I've been told that in Linux 2.6.35 if an application tries to write to a FLASH memory, but the FLASH is not ready, then the application's CPU is being "transferd" to the FLASH manipulation driver.
I've been also told that a patch was made to stop this behavior.

Can someone point me to this patch becuase I can't find it myself?

Thank you!
Hadar

[reginaldperrin]

Hmmm... not sure what is going on. Your description is a little vague.
However, from a hardware point of view, if the flash drive is USB, then the CPU handles read/writes anyway. This is not like the similar, but not so popular (Apple) Firewire, which has it's own chip in the port (and is therefore potentially faster).
If you see messages relating to CPU being used, this is possibly normal behaviour anyway, though it would seem not to be normal to display such a message.
Hope this helps.

[Rubberman]

Are you talking about a flash memory card such as a USB thumb drive or SD card? Or are you talking about system flash memory on the motherboard?

[thehadar]

I'm talking about a NAND flash device on the motherboard. My application writes log to a file, and I suspect that sometimes the write procedure takes precious time from my app's CPU slice.

Any feedback on this?

tnx

Are you talking about a flash memory card such as a USB thumb drive or SD card? Or are you talking about system flash memory on the motherboard?

[Rubberman]

I'm not familiar with this driver, but I suspect that it preempts your code and does not return to user space until the transfer is complete, a synchronous operation, rather than a non-blocking one. You might want to look on the kernel.org web site, or post the question to the Linux Foundation forums.