Trying not to top post.. Apologies before hand on my client restrictions. Anyways..
Manjunath, Yes, the call sequences are common. We may consider using cpu_relax() in the context of dvfs calls.. Except it could result in race connditions not in intrest. Do me a favor and flag the udelays u'd like to convert and send a patch for fix pls. Regards, Nishanth Menon ----- Original Message ----- From: G, Manjunath Kondaiah To: Menon, Nishanth; [email protected] <[email protected]> Cc: Imberton Guilhem <[email protected]>; Mike Chan <[email protected]>; Nayak, Rajendra; Roger Quadros <[email protected]>; Kalle Jokiniemi <[email protected]>; Reddy, Teerth; Kevin Hilman <[email protected]>; Paul Walmsley <[email protected]>; Hogander Jouni <[email protected]> Sent: Mon Oct 26 10:09:07 2009 Subject: RE: [PATCH 2/2 v3] OMAP3: PM: SR: SmartReflex Refactor Rev4.0 > -----Original Message----- > From: Menon, Nishanth > Sent: Monday, October 26, 2009 3:48 PM > To: G, Manjunath Kondaiah; [email protected] > Cc: Imberton Guilhem; Mike Chan; Nayak, Rajendra; Roger > Quadros; Kalle Jokiniemi; Reddy, Teerth; Kevin Hilman; Paul > Walmsley; Hogander Jouni > Subject: RE: [PATCH 2/2 v3] OMAP3: PM: SR: SmartReflex Refactor Rev4.0 > > > -----Original Message----- > > From: G, Manjunath Kondaiah > > Sent: Monday, October 26, 2009 3:40 AM > > To: Menon, Nishanth; [email protected] > > Cc: Imberton Guilhem; Mike Chan; Nayak, Rajendra; Roger > Quadros; Kalle > > Jokiniemi; Reddy, Teerth; Kevin Hilman; Paul Walmsley; > Hogander Jouni > > Subject: RE: [PATCH 2/2 v3] OMAP3: PM: SR: SmartReflex > Refactor Rev4.0 > > > > > > As per your comments for other patches when ever there is > udelay usage, > > cpu_relax is the better option. I see there are lot of > udelay(...) calls > > used in this patch. Why can't you use cpu_relax() or schedule(). > > Any specific reason? > > > Don’t really want to do cpu_relax in irq_locked context.. if > you look at the code flow, the call from cpu_idle is in > irq_locked.. Further this delay is part of bring up form > saved context where there is nothing else scheduled + we > don’t want anything else scheduled (and causing a change in > scheduling decision). So unfortunately, unlike standard > drivers, this cannot use the same reasoning. > NAK. My understanding is that, SR functions will be called during voltage scaling also. Voltage scaling will happen in non IRQ locked context. Please clarify if I am wrong. -Manjunath
