kevin/linux
1
Fork 0
Code
e51e2528d5
linux/drivers/ide/pci/cmd64x.c

750 lines
21 KiB
C
Raw Normal View History

cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
/*
cmd64x: interrupt status fixes (take 2) The driver's ide_dma_test_irq() method was reading the MRDMODE register even on PCI0643/6 where it was write-only -- fix this by always reading the "backward- compatible" interrupt bits, renaming dma_alt_stat to irq_stat as the interrupt status bits are not coupled to DMA. In addition, wrong interrupt bit was tested/cleared for the primary channel -- it's bit 2 in all the chip specs and the driver used bit 1... :-/ Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
* linux/drivers/ide/pci/cmd64x.c Version 1.44 Mar 12, 2007
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
*
* cmd64x.c: Enable interrupts at initialization time on Ultra/PCI machines.
* Due to massive hardware bugs, UltraDMA is only supported
* on the 646U2 and not on the 646U.
*
* Copyright (C) 1998 Eddie C. Dost (ecd@skynet.be)
* Copyright (C) 1998 David S. Miller (davem@redhat.com)
*
* Copyright (C) 1999-2002 Andre Hedrick <andre@linux-ide.org>
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
* Copyright (C) 2007 MontaVista Software, Inc. <source@mvista.com>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
*/
#include <linux/module.h>
#include <linux/types.h>
#include <linux/pci.h>
#include <linux/delay.h>
#include <linux/hdreg.h>
#include <linux/ide.h>
#include <linux/init.h>
#include <asm/io.h>
#define DISPLAY_CMD64X_TIMINGS
#define CMD_DEBUG 0
#if CMD_DEBUG
#define cmdprintk(x...) printk(x)
#else
#define cmdprintk(x...)
#endif
/*
* CMD64x specific registers definition.
*/
#define CFR 0x50
cmd64x: interrupt status fixes (take 2) The driver's ide_dma_test_irq() method was reading the MRDMODE register even on PCI0643/6 where it was write-only -- fix this by always reading the "backward- compatible" interrupt bits, renaming dma_alt_stat to irq_stat as the interrupt status bits are not coupled to DMA. In addition, wrong interrupt bit was tested/cleared for the primary channel -- it's bit 2 in all the chip specs and the driver used bit 1... :-/ Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
#define CFR_INTR_CH0 0x04
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
#define CNTRL 0x51
#define CNTRL_DIS_RA0 0x40
#define CNTRL_DIS_RA1 0x80
#define CNTRL_ENA_2ND 0x08
#define CMDTIM 0x52
#define ARTTIM0 0x53
#define DRWTIM0 0x54
#define ARTTIM1 0x55
#define DRWTIM1 0x56
#define ARTTIM23 0x57
#define ARTTIM23_DIS_RA2 0x04
#define ARTTIM23_DIS_RA3 0x08
#define ARTTIM23_INTR_CH1 0x10
#define ARTTIM2 0x57
#define ARTTIM3 0x57
#define DRWTIM23 0x58
#define DRWTIM2 0x58
#define BRST 0x59
#define DRWTIM3 0x5b
#define BMIDECR0 0x70
#define MRDMODE 0x71
#define MRDMODE_INTR_CH0 0x04
#define MRDMODE_INTR_CH1 0x08
#define MRDMODE_BLK_CH0 0x10
#define MRDMODE_BLK_CH1 0x20
#define BMIDESR0 0x72
#define UDIDETCR0 0x73
#define DTPR0 0x74
#define BMIDECR1 0x78
#define BMIDECSR 0x79
#define BMIDESR1 0x7A
#define UDIDETCR1 0x7B
#define DTPR1 0x7C
#if defined(DISPLAY_CMD64X_TIMINGS) && defined(CONFIG_PROC_FS)
#include <linux/stat.h>
#include <linux/proc_fs.h>
static u8 cmd64x_proc = 0;
#define CMD_MAX_DEVS 5
static struct pci_dev *cmd_devs[CMD_MAX_DEVS];
static int n_cmd_devs;
static char * print_cmd64x_get_info (char *buf, struct pci_dev *dev, int index)
{
char *p = buf;
u8 reg53 = 0, reg54 = 0, reg55 = 0, reg56 = 0; /* primary */
u8 reg57 = 0, reg58 = 0, reg5b; /* secondary */
u8 reg72 = 0, reg73 = 0; /* primary */
u8 reg7a = 0, reg7b = 0; /* secondary */
u8 reg50 = 0, reg71 = 0; /* extra */
p += sprintf(p, "\nController: %d\n", index);
p += sprintf(p, "CMD%x Chipset.\n", dev->device);
(void) pci_read_config_byte(dev, CFR, &reg50);
(void) pci_read_config_byte(dev, ARTTIM0, &reg53);
(void) pci_read_config_byte(dev, DRWTIM0, &reg54);
(void) pci_read_config_byte(dev, ARTTIM1, &reg55);
(void) pci_read_config_byte(dev, DRWTIM1, &reg56);
(void) pci_read_config_byte(dev, ARTTIM2, &reg57);
(void) pci_read_config_byte(dev, DRWTIM2, &reg58);
(void) pci_read_config_byte(dev, DRWTIM3, &reg5b);
(void) pci_read_config_byte(dev, MRDMODE, &reg71);
(void) pci_read_config_byte(dev, BMIDESR0, &reg72);
(void) pci_read_config_byte(dev, UDIDETCR0, &reg73);
(void) pci_read_config_byte(dev, BMIDESR1, &reg7a);
(void) pci_read_config_byte(dev, UDIDETCR1, &reg7b);
p += sprintf(p, "--------------- Primary Channel "
"---------------- Secondary Channel "
"-------------\n");
p += sprintf(p, " %sabled "
" %sabled\n",
(reg72&0x80)?"dis":" en",
(reg7a&0x80)?"dis":" en");
p += sprintf(p, "--------------- drive0 "
"--------- drive1 -------- drive0 "
"---------- drive1 ------\n");
p += sprintf(p, "DMA enabled: %s %s"
" %s %s\n",
(reg72&0x20)?"yes":"no ", (reg72&0x40)?"yes":"no ",
(reg7a&0x20)?"yes":"no ", (reg7a&0x40)?"yes":"no ");
p += sprintf(p, "DMA Mode: %s(%s) %s(%s)",
(reg72&0x20)?((reg73&0x01)?"UDMA":" DMA"):" PIO",
(reg72&0x20)?(
((reg73&0x30)==0x30)?(((reg73&0x35)==0x35)?"3":"0"):
((reg73&0x20)==0x20)?(((reg73&0x25)==0x25)?"3":"1"):
((reg73&0x10)==0x10)?(((reg73&0x15)==0x15)?"4":"2"):
((reg73&0x00)==0x00)?(((reg73&0x05)==0x05)?"5":"2"):
"X"):"?",
(reg72&0x40)?((reg73&0x02)?"UDMA":" DMA"):" PIO",
(reg72&0x40)?(
((reg73&0xC0)==0xC0)?(((reg73&0xC5)==0xC5)?"3":"0"):
((reg73&0x80)==0x80)?(((reg73&0x85)==0x85)?"3":"1"):
((reg73&0x40)==0x40)?(((reg73&0x4A)==0x4A)?"4":"2"):
((reg73&0x00)==0x00)?(((reg73&0x0A)==0x0A)?"5":"2"):
"X"):"?");
p += sprintf(p, " %s(%s) %s(%s)\n",
(reg7a&0x20)?((reg7b&0x01)?"UDMA":" DMA"):" PIO",
(reg7a&0x20)?(
((reg7b&0x30)==0x30)?(((reg7b&0x35)==0x35)?"3":"0"):
((reg7b&0x20)==0x20)?(((reg7b&0x25)==0x25)?"3":"1"):
((reg7b&0x10)==0x10)?(((reg7b&0x15)==0x15)?"4":"2"):
((reg7b&0x00)==0x00)?(((reg7b&0x05)==0x05)?"5":"2"):
"X"):"?",
(reg7a&0x40)?((reg7b&0x02)?"UDMA":" DMA"):" PIO",
(reg7a&0x40)?(
((reg7b&0xC0)==0xC0)?(((reg7b&0xC5)==0xC5)?"3":"0"):
((reg7b&0x80)==0x80)?(((reg7b&0x85)==0x85)?"3":"1"):
((reg7b&0x40)==0x40)?(((reg7b&0x4A)==0x4A)?"4":"2"):
((reg7b&0x00)==0x00)?(((reg7b&0x0A)==0x0A)?"5":"2"):
"X"):"?" );
p += sprintf(p, "PIO Mode: %s %s"
" %s %s\n",
"?", "?", "?", "?");
p += sprintf(p, " %s %s\n",
(reg50 & CFR_INTR_CH0) ? "interrupting" : "polling ",
(reg57 & ARTTIM23_INTR_CH1) ? "interrupting" : "polling");
p += sprintf(p, " %s %s\n",
(reg71 & MRDMODE_INTR_CH0) ? "pending" : "clear ",
(reg71 & MRDMODE_INTR_CH1) ? "pending" : "clear");
p += sprintf(p, " %s %s\n",
(reg71 & MRDMODE_BLK_CH0) ? "blocked" : "enabled",
(reg71 & MRDMODE_BLK_CH1) ? "blocked" : "enabled");
return (char *)p;
}
static int cmd64x_get_info (char *buffer, char **addr, off_t offset, int count)
{
char *p = buffer;
int i;
p += sprintf(p, "\n");
for (i = 0; i < n_cmd_devs; i++) {
struct pci_dev *dev = cmd_devs[i];
p = print_cmd64x_get_info(p, dev, i);
}
return p-buffer; /* => must be less than 4k! */
}
#endif /* defined(DISPLAY_CMD64X_TIMINGS) && defined(CONFIG_PROC_FS) */
cmd64x: fix recovery time calculation (take 3) The driver wrongly takes the address setup time into account when calculating the PIO recovery time -- this leads to slight overclocking of the PIO modes 0 and 1 (so, the prayers failed to help, as usual :-). Rework the code to be calculating recovery clock count as a difference between the total cycle count and the active count (we don't need to calculate the recovery time itself since it's not specified for the PIO modes 0 to 2, and for modes 3 and 4 this formula gives enough recovery time anyway in the chip's supported PCI frequency range). This patch has been inspired by reading the datasheets and looking at what the libata driver does; it has been compile-tested only (as usual :-) but anyway, the new code gives the same or longer recovery times than the old one... Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-17 13:57:24 -07:00
static u8 quantize_timing(int timing, int quant)
{
return (timing + quant - 1) / quant;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/*
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
* This routine calculates active/recovery counts and then writes them into
* the chipset registers.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
*/
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
static void program_cycle_times (ide_drive_t *drive, int cycle_time, int active_time)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
{
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
struct pci_dev *dev = HWIF(drive)->pci_dev;
int clock_time = 1000 / system_bus_clock();
u8 cycle_count, active_count, recovery_count, drwtim;
static const u8 recovery_values[] =
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
{15, 15, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 0};
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
static const u8 drwtim_regs[4] = {DRWTIM0, DRWTIM1, DRWTIM2, DRWTIM3};
cmdprintk("program_cycle_times parameters: total=%d, active=%d\n",
cycle_time, active_time);
cycle_count = quantize_timing( cycle_time, clock_time);
active_count = quantize_timing(active_time, clock_time);
recovery_count = cycle_count - active_count;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/*
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
* In case we've got too long recovery phase, try to lengthen
* the active phase
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
*/
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
if (recovery_count > 16) {
active_count += recovery_count - 16;
recovery_count = 16;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
if (active_count > 16) /* shouldn't actually happen... */
active_count = 16;
cmdprintk("Final counts: total=%d, active=%d, recovery=%d\n",
cycle_count, active_count, recovery_count);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/*
* Convert values to internal chipset representation
*/
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
recovery_count = recovery_values[recovery_count];
active_count &= 0x0f;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
/* Program the active/recovery counts into the DRWTIM register */
drwtim = (active_count << 4) | recovery_count;
(void) pci_write_config_byte(dev, drwtim_regs[drive->dn], drwtim);
cmdprintk("Write 0x%02x to reg 0x%x\n", drwtim, drwtim_regs[drive->dn]);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
/*
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
* This routine selects drive's best PIO mode and writes into the chipset
* registers setup/active/recovery timings.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
*/
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
static u8 cmd64x_tune_pio (ide_drive_t *drive, u8 mode_wanted)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
{
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
ide_hwif_t *hwif = HWIF(drive);
struct pci_dev *dev = hwif->pci_dev;
cmd64x: fix recovery time calculation (take 3) The driver wrongly takes the address setup time into account when calculating the PIO recovery time -- this leads to slight overclocking of the PIO modes 0 and 1 (so, the prayers failed to help, as usual :-). Rework the code to be calculating recovery clock count as a difference between the total cycle count and the active count (we don't need to calculate the recovery time itself since it's not specified for the PIO modes 0 to 2, and for modes 3 and 4 this formula gives enough recovery time anyway in the chip's supported PCI frequency range). This patch has been inspired by reading the datasheets and looking at what the libata driver does; it has been compile-tested only (as usual :-) but anyway, the new code gives the same or longer recovery times than the old one... Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-17 13:57:24 -07:00
ide_pio_data_t pio;
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
u8 pio_mode, setup_count, arttim = 0;
static const u8 setup_values[] = {0x40, 0x40, 0x40, 0x80, 0, 0xc0};
static const u8 arttim_regs[4] = {ARTTIM0, ARTTIM1, ARTTIM23, ARTTIM23};
cmd64x: fix recovery time calculation (take 3) The driver wrongly takes the address setup time into account when calculating the PIO recovery time -- this leads to slight overclocking of the PIO modes 0 and 1 (so, the prayers failed to help, as usual :-). Rework the code to be calculating recovery clock count as a difference between the total cycle count and the active count (we don't need to calculate the recovery time itself since it's not specified for the PIO modes 0 to 2, and for modes 3 and 4 this formula gives enough recovery time anyway in the chip's supported PCI frequency range). This patch has been inspired by reading the datasheets and looking at what the libata driver does; it has been compile-tested only (as usual :-) but anyway, the new code gives the same or longer recovery times than the old one... Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-17 13:57:24 -07:00
pio_mode = ide_get_best_pio_mode(drive, mode_wanted, 5, &pio);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
cmdprintk("%s: PIO mode wanted %d, selected %d (%d ns)%s\n",
drive->name, mode_wanted, pio_mode, pio.cycle_time,
pio.overridden ? " (overriding vendor mode)" : "");
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
program_cycle_times(drive, pio.cycle_time,
ide_pio_timings[pio_mode].active_time);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
setup_count = quantize_timing(ide_pio_timings[pio_mode].setup_time,
1000 / system_bus_clock());
/*
* The primary channel has individual address setup timing registers
* for each drive and the hardware selects the slowest timing itself.
* The secondary channel has one common register and we have to select
* the slowest address setup timing ourselves.
*/
if (hwif->channel) {
ide_drive_t *drives = hwif->drives;
drive->drive_data = setup_count;
setup_count = max(drives[0].drive_data, drives[1].drive_data);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
if (setup_count > 5) /* shouldn't actually happen... */
setup_count = 5;
cmdprintk("Final address setup count: %d\n", setup_count);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
/*
* Program the address setup clocks into the ARTTIM registers.
* Avoid clearing the secondary channel's interrupt bit.
*/
(void) pci_read_config_byte (dev, arttim_regs[drive->dn], &arttim);
if (hwif->channel)
arttim &= ~ARTTIM23_INTR_CH1;
arttim &= ~0xc0;
arttim |= setup_values[setup_count];
(void) pci_write_config_byte(dev, arttim_regs[drive->dn], arttim);
cmdprintk("Write 0x%02x to reg 0x%x\n", arttim, arttim_regs[drive->dn]);
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
return pio_mode;
}
/*
* Attempts to set drive's PIO mode.
* Special cases are 8: prefetch off, 9: prefetch on (both never worked),
* and 255: auto-select best mode (used at boot time).
*/
static void cmd64x_tune_drive (ide_drive_t *drive, u8 pio)
{
/*
* Filter out the prefetch control values
* to prevent PIO5 from being programmed
*/
if (pio == 8 || pio == 9)
return;
pio = cmd64x_tune_pio(drive, pio);
(void) ide_config_drive_speed(drive, XFER_PIO_0 + pio);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
static u8 cmd64x_ratemask (ide_drive_t *drive)
{
struct pci_dev *dev = HWIF(drive)->pci_dev;
u8 mode = 0;
switch(dev->device) {
case PCI_DEVICE_ID_CMD_649:
mode = 3;
break;
case PCI_DEVICE_ID_CMD_648:
mode = 2;
break;
case PCI_DEVICE_ID_CMD_643:
return 0;
case PCI_DEVICE_ID_CMD_646:
{
unsigned int class_rev = 0;
pci_read_config_dword(dev,
PCI_CLASS_REVISION, &class_rev);
class_rev &= 0xff;
/*
* UltraDMA only supported on PCI646U and PCI646U2, which
* correspond to revisions 0x03, 0x05 and 0x07 respectively.
* Actually, although the CMD tech support people won't
* tell me the details, the 0x03 revision cannot support
* UDMA correctly without hardware modifications, and even
* then it only works with Quantum disks due to some
* hold time assumptions in the 646U part which are fixed
* in the 646U2.
*
* So we only do UltraDMA on revision 0x05 and 0x07 chipsets.
*/
switch(class_rev) {
case 0x07:
case 0x05:
return 1;
case 0x03:
case 0x01:
default:
return 0;
}
}
}
if (!eighty_ninty_three(drive))
mode = min(mode, (u8)1);
return mode;
}
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
static int cmd64x_tune_chipset (ide_drive_t *drive, u8 speed)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
{
ide_hwif_t *hwif = HWIF(drive);
struct pci_dev *dev = hwif->pci_dev;
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
u8 unit = drive->dn & 0x01;
u8 regU = 0, pciU = hwif->channel ? UDIDETCR1 : UDIDETCR0;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
speed = ide_rate_filter(cmd64x_ratemask(drive), speed);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
if (speed >= XFER_SW_DMA_0) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
(void) pci_read_config_byte(dev, pciU, &regU);
regU &= ~(unit ? 0xCA : 0x35);
}
switch(speed) {
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
case XFER_UDMA_5:
regU |= unit ? 0x0A : 0x05;
break;
case XFER_UDMA_4:
regU |= unit ? 0x4A : 0x15;
break;
case XFER_UDMA_3:
regU |= unit ? 0x8A : 0x25;
break;
case XFER_UDMA_2:
regU |= unit ? 0x42 : 0x11;
break;
case XFER_UDMA_1:
regU |= unit ? 0x82 : 0x21;
break;
case XFER_UDMA_0:
regU |= unit ? 0xC2 : 0x31;
break;
case XFER_MW_DMA_2:
program_cycle_times(drive, 120, 70);
break;
case XFER_MW_DMA_1:
program_cycle_times(drive, 150, 80);
break;
case XFER_MW_DMA_0:
program_cycle_times(drive, 480, 215);
break;
case XFER_PIO_5:
case XFER_PIO_4:
case XFER_PIO_3:
case XFER_PIO_2:
case XFER_PIO_1:
case XFER_PIO_0:
(void) cmd64x_tune_pio(drive, speed - XFER_PIO_0);
break;
default:
return 1;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
if (speed >= XFER_SW_DMA_0)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
(void) pci_write_config_byte(dev, pciU, regU);
cmd64x: fix multiword and remove single-word DMA support Fix the multiword DMA and drop the single-word DMA support (which nobody will miss, I think). In order to do it, a number of changes was necessary: - rename program_drive_counts() to program_cycle_times(), pass to it cycle's total/active times instead of the clock counts, and convert them into the active/recovery clocks there instead of cmd64x_tune_pio() -- this causes quantize_timing() to also move; - contrarywise, move all the code handling the address setup timing into cmd64x_tune_pio(), so that setting MWDMA mode wouldn't change address setup; - remove from the speedproc() method the bogus code pretending to set the DMA timings by twiddling bits in the BMIDE status register, handle setting MWDMA by just calling program_cycle_times(); while at it, improve the style of that whole switch statement; - stop fiddling with the DMA capable bits in the speedproc() method -- they do not enable DMA, and are properly dealt with by the dma_host_{on,off} methods; - don't set hwif->swdma_mask in the init_hwif() method anymore. In addition to those changes, do the following: - in cmd64x_tune_pio(), when writing to ARTTIM23 register preserve the interrupt status bit, eliminate local_irq_{save|restore}() around this code as there's *no* actual race with the interrupt handler, and move cmdprintk() to a more fitting place -- after ide_get_best_pio_mode() call; - make {arttim|drwtim}_regs arrays single-dimensional, indexed with drive->dn; - rename {setup|recovery}_counts[] into more fitting {setup|recovery}_values[]; - in the speedproc() method, get rid of the duplicate reads/writes from/to the UDIDETCRx registers and of the extra variable used to store the transfer mode value after filtering, use another method of determining master/slave drive, and cleanup useless parens; - beautify cmdprintk() output here and there. While at it, remove meaningless comment about the driver being used only on UltraSPARC and long non-relevant RCS tag. :-) Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
return ide_config_drive_speed(drive, speed);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
static int config_chipset_for_dma (ide_drive_t *drive)
{
u8 speed = ide_dma_speed(drive, cmd64x_ratemask(drive));
if (!speed)
return 0;
ide: unexport ide_set_xfer_rate() (v2) In cmd64x, siimage and scc_pata drivers: * don't set drive->init_speed as it should be already set by successful execution of ide_set_xfer_rate() * use hwif->speedproc functions directly Above changes allows removal of EXPORT_SYMBOL_GPL(ide_set_xfer_rate). v2: * updated for scc_pata driver Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:24 -07:00
if (cmd64x_tune_chipset(drive, speed))
return 0;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
return ide_dma_enable(drive);
}
static int cmd64x_config_drive_for_dma (ide_drive_t *drive)
{
ide: add ide_use_fast_pio() helper (v3) * add ide_use_fast_pio() helper for use by host drivers * add DMA capability and hwif->autodma checks to ide_use_dma() - au1xxx-ide/it8213/it821x drivers didn't check for (id->capability & 1) [ for the IT8211/2 in SMART mode this check shouldn't be made but since in it821x_fixups() we set DMA bit explicitly: if(strstr(id->model, "Integrated Technology Express")) { /* In raid mode the ident block is slightly buggy We need to set the bits so that the IDE layer knows LBA28. LBA48 and DMA ar valid */ id->capability |= 3; /* LBA28, DMA */ we are better off using generic helper if we can ] - ide-cris driver didn't set ->autodma [ before the patch hwif->autodma was only checked in the chipset specific hwif->ide_dma_check implementations, for ide-cris it is cris_dma_check() function so there no behavior change here ] v2: * updated patch description (thanks to Alan Cox for the feedback) v3: * updated for scc_pata driver Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
if (ide_use_dma(drive) && config_chipset_for_dma(drive))
ide: add ide_set_dma() helper (v2) * add ide_set_dma() helper and make ide_hwif_t.ide_dma_check return -1 when DMA needs to be disabled (== need to call ->ide_dma_off_quietly) 0 when DMA needs to be enabled (== need to call ->ide_dma_on) 1 when DMA setting shouldn't be changed * fix IDE code to use ide_set_dma() instead if using ->ide_dma_check directly v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:26 -07:00
return 0;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
ide: disable DMA in ->ide_dma_check for "no IORDY" case (v2) If DMA is unsupported ->ide_dma_check should disable DMA. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
if (ide_use_fast_pio(drive))
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
cmd64x_tune_drive(drive, 255);
ide: disable DMA in ->ide_dma_check for "no IORDY" case (v2) If DMA is unsupported ->ide_dma_check should disable DMA. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
ide: add ide_set_dma() helper (v2) * add ide_set_dma() helper and make ide_hwif_t.ide_dma_check return -1 when DMA needs to be disabled (== need to call ->ide_dma_off_quietly) 0 when DMA needs to be enabled (== need to call ->ide_dma_on) 1 when DMA setting shouldn't be changed * fix IDE code to use ide_set_dma() instead if using ->ide_dma_check directly v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:26 -07:00
return -1;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
}
static int cmd64x_alt_dma_status (struct pci_dev *dev)
{
switch(dev->device) {
case PCI_DEVICE_ID_CMD_648:
case PCI_DEVICE_ID_CMD_649:
return 1;
default:
break;
}
return 0;
}
static int cmd64x_ide_dma_end (ide_drive_t *drive)
{
u8 dma_stat = 0, dma_cmd = 0;
ide_hwif_t *hwif = HWIF(drive);
struct pci_dev *dev = hwif->pci_dev;
drive->waiting_for_dma = 0;
/* read DMA command state */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
dma_cmd = inb(hwif->dma_command);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* stop DMA */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
outb(dma_cmd & ~1, hwif->dma_command);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* get DMA status */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
dma_stat = inb(hwif->dma_status);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* clear the INTR & ERROR bits */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
outb(dma_stat | 6, hwif->dma_status);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
if (cmd64x_alt_dma_status(dev)) {
u8 dma_intr = 0;
u8 dma_mask = (hwif->channel) ? ARTTIM23_INTR_CH1 :
CFR_INTR_CH0;
u8 dma_reg = (hwif->channel) ? ARTTIM2 : CFR;
(void) pci_read_config_byte(dev, dma_reg, &dma_intr);
/* clear the INTR bit */
(void) pci_write_config_byte(dev, dma_reg, dma_intr|dma_mask);
}
/* purge DMA mappings */
ide_destroy_dmatable(drive);
/* verify good DMA status */
return (dma_stat & 7) != 4;
}
static int cmd64x_ide_dma_test_irq (ide_drive_t *drive)
{
cmd64x: interrupt status fixes (take 2) The driver's ide_dma_test_irq() method was reading the MRDMODE register even on PCI0643/6 where it was write-only -- fix this by always reading the "backward- compatible" interrupt bits, renaming dma_alt_stat to irq_stat as the interrupt status bits are not coupled to DMA. In addition, wrong interrupt bit was tested/cleared for the primary channel -- it's bit 2 in all the chip specs and the driver used bit 1... :-/ Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
ide_hwif_t *hwif = HWIF(drive);
struct pci_dev *dev = hwif->pci_dev;
u8 irq_reg = hwif->channel ? ARTTIM23 : CFR;
u8 irq_stat = 0, mask = hwif->channel ? ARTTIM23_INTR_CH1 : CFR_INTR_CH0;
u8 dma_stat = inb(hwif->dma_status);
(void) pci_read_config_byte(dev, irq_reg, &irq_stat);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
#ifdef DEBUG
cmd64x: interrupt status fixes (take 2) The driver's ide_dma_test_irq() method was reading the MRDMODE register even on PCI0643/6 where it was write-only -- fix this by always reading the "backward- compatible" interrupt bits, renaming dma_alt_stat to irq_stat as the interrupt status bits are not coupled to DMA. In addition, wrong interrupt bit was tested/cleared for the primary channel -- it's bit 2 in all the chip specs and the driver used bit 1... :-/ Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
printk("%s: dma_stat: 0x%02x irq_stat: 0x%02x mask: 0x%02x\n",
drive->name, dma_stat, irq_stat, mask);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
#endif
cmd64x: interrupt status fixes (take 2) The driver's ide_dma_test_irq() method was reading the MRDMODE register even on PCI0643/6 where it was write-only -- fix this by always reading the "backward- compatible" interrupt bits, renaming dma_alt_stat to irq_stat as the interrupt status bits are not coupled to DMA. In addition, wrong interrupt bit was tested/cleared for the primary channel -- it's bit 2 in all the chip specs and the driver used bit 1... :-/ Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-05-05 13:03:49 -07:00
if (!(irq_stat & mask))
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
return 0;
/* return 1 if INTR asserted */
if ((dma_stat & 4) == 4)
return 1;
return 0;
}
/*
* ASUS P55T2P4D with CMD646 chipset revision 0x01 requires the old
* event order for DMA transfers.
*/
static int cmd646_1_ide_dma_end (ide_drive_t *drive)
{
ide_hwif_t *hwif = HWIF(drive);
u8 dma_stat = 0, dma_cmd = 0;
drive->waiting_for_dma = 0;
/* get DMA status */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
dma_stat = inb(hwif->dma_status);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* read DMA command state */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
dma_cmd = inb(hwif->dma_command);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* stop DMA */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
outb(dma_cmd & ~1, hwif->dma_command);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* clear the INTR & ERROR bits */
ide: use PIO/MMIO operations directly where possible (v2) This results in smaller/faster/simpler code and allows future optimizations. Also remove no longer needed ide[_mm]_{inl,outl}() and ide_hwif_t.{INL,OUTL}. v2: * updated for scc_pata Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-02-16 18:40:25 -07:00
outb(dma_stat | 6, hwif->dma_status);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
/* and free any DMA resources */
ide_destroy_dmatable(drive);
/* verify good DMA status */
return (dma_stat & 7) != 4;
}
static unsigned int __devinit init_chipset_cmd64x(struct pci_dev *dev, const char *name)
{
u32 class_rev = 0;
u8 mrdmode = 0;
pci_read_config_dword(dev, PCI_CLASS_REVISION, &class_rev);
class_rev &= 0xff;
switch(dev->device) {
case PCI_DEVICE_ID_CMD_643:
break;
case PCI_DEVICE_ID_CMD_646:
printk(KERN_INFO "%s: chipset revision 0x%02X, ", name, class_rev);
switch(class_rev) {
case 0x07:
case 0x05:
printk("UltraDMA Capable");
break;
case 0x03:
printk("MultiWord DMA Force Limited");
break;
case 0x01:
default:
printk("MultiWord DMA Limited, IRQ workaround enabled");
break;
}
printk("\n");
break;
case PCI_DEVICE_ID_CMD_648:
case PCI_DEVICE_ID_CMD_649:
break;
default:
break;
}
/* Set a good latency timer and cache line size value. */
(void) pci_write_config_byte(dev, PCI_LATENCY_TIMER, 64);
/* FIXME: pci_set_master() to ensure a good latency timer value */
/* Setup interrupts. */
(void) pci_read_config_byte(dev, MRDMODE, &mrdmode);
mrdmode &= ~(0x30);
(void) pci_write_config_byte(dev, MRDMODE, mrdmode);
/* Use MEMORY READ LINE for reads.
* NOTE: Although not mentioned in the PCI0646U specs,
* these bits are write only and won't be read
* back as set or not. The PCI0646U2 specs clarify
* this point.
*/
(void) pci_write_config_byte(dev, MRDMODE, mrdmode | 0x02);
/* Set reasonable active/recovery/address-setup values. */
(void) pci_write_config_byte(dev, ARTTIM0, 0x40);
(void) pci_write_config_byte(dev, DRWTIM0, 0x3f);
(void) pci_write_config_byte(dev, ARTTIM1, 0x40);
(void) pci_write_config_byte(dev, DRWTIM1, 0x3f);
#ifdef __i386__
(void) pci_write_config_byte(dev, ARTTIM23, 0x1c);
#else
(void) pci_write_config_byte(dev, ARTTIM23, 0x5c);
#endif
(void) pci_write_config_byte(dev, DRWTIM23, 0x3f);
(void) pci_write_config_byte(dev, DRWTIM3, 0x3f);
#ifdef CONFIG_PPC
(void) pci_write_config_byte(dev, UDIDETCR0, 0xf0);
#endif /* CONFIG_PPC */
#if defined(DISPLAY_CMD64X_TIMINGS) && defined(CONFIG_PROC_FS)
cmd_devs[n_cmd_devs++] = dev;
if (!cmd64x_proc) {
cmd64x_proc = 1;
ide_pci_create_host_proc("cmd64x", cmd64x_get_info);
}
#endif /* DISPLAY_CMD64X_TIMINGS && CONFIG_PROC_FS */
return 0;
}
static unsigned int __devinit ata66_cmd64x(ide_hwif_t *hwif)
{
u8 ata66 = 0, mask = (hwif->channel) ? 0x02 : 0x01;
switch(hwif->pci_dev->device) {
case PCI_DEVICE_ID_CMD_643:
case PCI_DEVICE_ID_CMD_646:
return ata66;
default:
break;
}
pci_read_config_byte(hwif->pci_dev, BMIDECSR, &ata66);
return (ata66 & mask) ? 1 : 0;
}
static void __devinit init_hwif_cmd64x(ide_hwif_t *hwif)
{
struct pci_dev *dev = hwif->pci_dev;
unsigned int class_rev;
hwif->autodma = 0;
pci_read_config_dword(dev, PCI_CLASS_REVISION, &class_rev);
class_rev &= 0xff;
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
hwif->tuneproc = &cmd64x_tune_drive;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
hwif->speedproc = &cmd64x_tune_chipset;
cmd64x: fix PIO mode setup (take 3) The driver's tuneproc() method fails to set the drive's own speed -- fix this by renaming the function to cmd64x_tune_pio(), making it return the mode set, and "wrapping" the new tuneproc() method around it; while at it, also get rid of the non-working prefetch control code (filtering out related argument values in the "wrapper"), remove redundant PIO5 mode limitation, make cmdprintk() give more sensible mode info, and remove mention about the obsolete /proc/ interface. Get rid of the broken config_chipset_for_pio() which always tried to set PIO4, switch to always auto-tuning PIO instead. Oh, and add the missing PIO5 support to the speedproc() method while at it. :-) Warning: compile tested only -- getting to the real hardware isn't that easy... On Tuesday 06 February 2007 22:11, Mikael Pettersson <mikpe@it.uu.se> wrote: > > Worked fine on my SPARC Ultra5 with a CMD646 IDE controller. Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-03-03 09:48:53 -07:00
hwif->drives[0].autotune = hwif->drives[1].autotune = 1;
if (!hwif->dma_base)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
return;
hwif->atapi_dma = 1;
hwif->ultra_mask = 0x3f;
hwif->mwdma_mask = 0x07;
if (dev->device == PCI_DEVICE_ID_CMD_643)
hwif->ultra_mask = 0x80;
if (dev->device == PCI_DEVICE_ID_CMD_646)
hwif->ultra_mask = (class_rev > 0x04) ? 0x07 : 0x80;
if (dev->device == PCI_DEVICE_ID_CMD_648)
hwif->ultra_mask = 0x1f;
hwif->ide_dma_check = &cmd64x_config_drive_for_dma;
if (!(hwif->udma_four))
hwif->udma_four = ata66_cmd64x(hwif);
if (dev->device == PCI_DEVICE_ID_CMD_646) {
hwif->chipset = ide_cmd646;
if (class_rev == 0x01) {
hwif->ide_dma_end = &cmd646_1_ide_dma_end;
} else {
hwif->ide_dma_end = &cmd64x_ide_dma_end;
hwif->ide_dma_test_irq = &cmd64x_ide_dma_test_irq;
}
} else {
hwif->ide_dma_end = &cmd64x_ide_dma_end;
hwif->ide_dma_test_irq = &cmd64x_ide_dma_test_irq;
}
if (!noautodma)
hwif->autodma = 1;
hwif->drives[0].autodma = hwif->autodma;
hwif->drives[1].autodma = hwif->autodma;
}
static ide_pci_device_t cmd64x_chipsets[] __devinitdata = {
{ /* 0 */
.name = "CMD643",
.init_chipset = init_chipset_cmd64x,
.init_hwif = init_hwif_cmd64x,
.channels = 2,
.autodma = AUTODMA,
.bootable = ON_BOARD,
},{ /* 1 */
.name = "CMD646",
.init_chipset = init_chipset_cmd64x,
.init_hwif = init_hwif_cmd64x,
.channels = 2,
.autodma = AUTODMA,
.enablebits = {{0x00,0x00,0x00}, {0x51,0x80,0x80}},
.bootable = ON_BOARD,
},{ /* 2 */
.name = "CMD648",
.init_chipset = init_chipset_cmd64x,
.init_hwif = init_hwif_cmd64x,
.channels = 2,
.autodma = AUTODMA,
.bootable = ON_BOARD,
},{ /* 3 */
.name = "CMD649",
.init_chipset = init_chipset_cmd64x,
.init_hwif = init_hwif_cmd64x,
.channels = 2,
.autodma = AUTODMA,
.bootable = ON_BOARD,
}
};
static int __devinit cmd64x_init_one(struct pci_dev *dev, const struct pci_device_id *id)
{
return ide_setup_pci_device(dev, &cmd64x_chipsets[id->driver_data]);
}
static struct pci_device_id cmd64x_pci_tbl[] = {
{ PCI_VENDOR_ID_CMD, PCI_DEVICE_ID_CMD_643, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
{ PCI_VENDOR_ID_CMD, PCI_DEVICE_ID_CMD_646, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 1},
{ PCI_VENDOR_ID_CMD, PCI_DEVICE_ID_CMD_648, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 2},
{ PCI_VENDOR_ID_CMD, PCI_DEVICE_ID_CMD_649, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 3},
{ 0, },
};
MODULE_DEVICE_TABLE(pci, cmd64x_pci_tbl);
static struct pci_driver driver = {
.name = "CMD64x_IDE",
.id_table = cmd64x_pci_tbl,
.probe = cmd64x_init_one,
};
ide: add missing __init tags to IDE PCI host drivers also change __devinit tag for sgiioc4.c:ioc4_ide_init() to __init Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2007-01-27 05:46:56 -07:00
static int __init cmd64x_ide_init(void)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 15:20:36 -07:00
{
return ide_pci_register_driver(&driver);
}
module_init(cmd64x_ide_init);
MODULE_AUTHOR("Eddie Dost, David Miller, Andre Hedrick");
MODULE_DESCRIPTION("PCI driver module for CMD64x IDE");
MODULE_LICENSE("GPL");
Copy Permalink