Is it dangerous using EKern::Sleep() in kernel extension?

Hi, I am a audio device driever engineer.=20
I use many EKern::Sleep() in codec kern extenstion for reduce codec pop noi=
se.=20
But somebody say "It is very dangerous".

I want know why it is sangerous and what is casuse by usging this function.=
=20

thank you.

Edited by: happypunk on 20-Aug-2008 13:18

Hi

If the sleep is a blocking sleep, e.g. waits for a counter to expire, then =
this can cause problems depending on the length of time you want to sleep.

Do you mean NKern::Sleep(TUint32 aTime) ?
Puts the current nanothread to sleep for the specified duration and it can =
be called from Symbian OS threads.

Mark
Symbian

oh, I mean NKern::Sleep().=20

Is not it dangerous?

NKern::Sleep will block the current thread for a number of milliseconds acc=
ording to the parameter you give.
Other threads can still run, so it is often safe.

Times when NKern::Sleep should not be used:
1) If you don't know what context the function is called in
2) If your driver is executing in DFC0, DFC1 or interrupt context
DFC0 is a shared thread, so all drivers using DFC0 are also blocked (a <=
5ms sleep is probably fine, but long sleeps can cause side effects in other=
drivers because their DFC runs too late)
DFC1 is the timer thread, calling NKern::Sleep here will deadlock the syste=
m
In interrupt context, the system will crash

Recommended context to use NKern::Sleep:
1) in the context of a user thread
For example when handling a DoControl
2) in your device driver's private DFC thread.
Private DFC threads, created with Kern:ynamicDfcQCreate (from 9.3) or Ker=
n:fcQCreate (from 8.1b) are recommended because your device driver can ma=
ke blocking calls with NKern::Sleep or fast mutex / fast semaphore without =
affecting other drivers.

PS: the very dangerous function is Kern::NanoWait, which is a spinning dela=
y loop and should only be used for microsecond size delays. (it blocks all =
lower priority threads in the system)