Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752532Ab0BNRsi (ORCPT ); Sun, 14 Feb 2010 12:48:38 -0500 Received: from einhorn.in-berlin.de ([192.109.42.8]:41366 "EHLO einhorn.in-berlin.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752388Ab0BNRsg (ORCPT ); Sun, 14 Feb 2010 12:48:36 -0500 X-Envelope-From: stefanr@s5r6.in-berlin.de Date: Sun, 14 Feb 2010 18:47:47 +0100 (CET) From: Stefan Richter Subject: [PATCH 3/4] firewire: get_cycle_timer optimization and cleanup To: linux-kernel@vger.kernel.org cc: linux1394-devel@lists.sourceforge.net In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; CHARSET=us-ascii Content-Disposition: INLINE Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6032 Lines: 174 ohci: Break out of the retry loop if too many attempts were necessary. This may theoretically happen if the chip is fatally defective or if the get_cycle_timer ioctl was performed after a CardBus controller was ejected. Also micro-optimize the loop by re-using the last two register reads in the next iteration, remove a questionable inline keyword, and shuffle a comment around. core: ioctl_get_cycle_timer() is always called with interrupts on, therefore local_irq_save() can be replaced by local_irq_disable(). Disabled local IRQs imply disabled preemption, hence preempt_disable() can be removed. Signed-off-by: Stefan Richter --- drivers/firewire/core-cdev.c | 17 ++++------ drivers/firewire/ohci.c | 57 ++++++++++++++++++----------------- 2 files changed, 36 insertions(+), 38 deletions(-) Index: linux-2.6.33-rc8/drivers/firewire/core-cdev.c =================================================================== --- linux-2.6.33-rc8.orig/drivers/firewire/core-cdev.c +++ linux-2.6.33-rc8/drivers/firewire/core-cdev.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include @@ -32,8 +33,6 @@ #include #include #include -#include -#include #include #include #include @@ -1035,21 +1034,19 @@ static int ioctl_get_cycle_timer(struct { struct fw_cdev_get_cycle_timer *request = buffer; struct fw_card *card = client->device->card; - unsigned long long bus_time; struct timeval tv; - unsigned long flags; + u32 cycle_time; - preempt_disable(); - local_irq_save(flags); + local_irq_disable(); - bus_time = card->driver->get_bus_time(card); + cycle_time = card->driver->get_bus_time(card); do_gettimeofday(&tv); - local_irq_restore(flags); - preempt_enable(); + local_irq_enable(); request->local_time = tv.tv_sec * 1000000ULL + tv.tv_usec; - request->cycle_timer = bus_time & 0xffffffff; + request->cycle_timer = cycle_time; + return 0; } Index: linux-2.6.33-rc8/drivers/firewire/ohci.c =================================================================== --- linux-2.6.33-rc8.orig/drivers/firewire/ohci.c +++ linux-2.6.33-rc8/drivers/firewire/ohci.c @@ -1795,60 +1795,61 @@ static int ohci_enable_phys_dma(struct f #endif /* CONFIG_FIREWIRE_OHCI_REMOTE_DMA */ } -static inline u32 cycle_timer_ticks(u32 cycle_timer) +static u32 cycle_timer_ticks(u32 cycle_timer) { u32 ticks; ticks = cycle_timer & 0xfff; ticks += 3072 * ((cycle_timer >> 12) & 0x1fff); ticks += (3072 * 8000) * (cycle_timer >> 25); + return ticks; } +/* + * Some controllers exhibit one or more of the following bugs when updating the + * iso cycle timer register: + * - When the lowest six bits are wrapping around to zero, a read that happens + * at the same time will return garbage in the lowest ten bits. + * - When the cycleOffset field wraps around to zero, the cycleCount field is + * not incremented for about 60 ns. + * - Occasionally, the entire register reads zero. + * + * To catch these, we read the register three times and ensure that the + * difference between each two consecutive reads is approximately the same, i.e. + * less than twice the other. Furthermore, any negative difference indicates an + * error. (A PCI read should take at least 20 ticks of the 24.576 MHz timer to + * execute, so we have enough precision to compute the ratio of the differences.) + */ static u64 ohci_get_bus_time(struct fw_card *card) { struct fw_ohci *ohci = fw_ohci(card); u32 c0, c1, c2; u32 t0, t1, t2; s32 diff01, diff12; - u64 bus_time; + int i; + + c2 = reg_read(ohci, OHCI1394_IsochronousCycleTimer); - if (!ohci->iso_cycle_timer_quirk) { + if (ohci->iso_cycle_timer_quirk) { + i = 0; + c1 = c2; c2 = reg_read(ohci, OHCI1394_IsochronousCycleTimer); - } else { - /* - * Some controllers exhibit one or more of the following bugs - * when updating the iso cycle timer register: - * - When the lowest six bits are wrapping around to zero, - * a read that happens at the same time will return garbage - * in the lowest ten bits. - * - When the cycleOffset field wraps around to zero, the - * cycleCount field is not incremented for about 60 ns. - * - Occasionally, the entire register reads zero. - * - * To catch these, we read the register three times and ensure - * that the difference between each two consecutive reads is - * approximately the same, i.e., less than twice the other. - * Furthermore, any negative difference indicates an error. - * (A PCI read should take at least 20 ticks of the 24.576 MHz - * timer to execute, so we have enough precision to compute the - * ratio of the differences.) - */ do { - c0 = reg_read(ohci, OHCI1394_IsochronousCycleTimer); - c1 = reg_read(ohci, OHCI1394_IsochronousCycleTimer); + c0 = c1; + c1 = c2; c2 = reg_read(ohci, OHCI1394_IsochronousCycleTimer); t0 = cycle_timer_ticks(c0); t1 = cycle_timer_ticks(c1); t2 = cycle_timer_ticks(c2); diff01 = t1 - t0; diff12 = t2 - t1; - } while (diff01 <= 0 || diff12 <= 0 || - diff01 / diff12 >= 2 || diff12 / diff01 >= 2); + } while ((diff01 <= 0 || diff12 <= 0 || + diff01 / diff12 >= 2 || diff12 / diff01 >= 2) + && i++ < 20); } - bus_time = ((u64)atomic_read(&ohci->bus_seconds) << 32) | c2; - return bus_time; + return ((u64)atomic_read(&ohci->bus_seconds) << 32) | c2; } static void copy_iso_headers(struct iso_context *ctx, void *p) -- Stefan Richter -=====-==-=- --=- -===- http://arcgraph.de/sr/ -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/