rockbox/firmware/drivers/ata.c

1420 lines
35 KiB
C
Raw Normal View History

/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
* Copyright (C) 2002 by Alan Korr
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
****************************************************************************/
#include <stdbool.h>
#include <inttypes.h>
#include "led.h"
#include "cpu.h"
#include "system.h"
#include "debug.h"
#include "panic.h"
#include "power.h"
#include "string.h"
#include "ata-driver.h"
#include "ata-defines.h"
#include "fs_defines.h"
#include "storage.h"
#define SELECT_DEVICE1 0x10
#define SELECT_LBA 0x40
#define CONTROL_nIEN 0x02
#define CONTROL_SRST 0x04
#define CMD_READ_SECTORS 0x20
#define CMD_WRITE_SECTORS 0x30
#define CMD_WRITE_SECTORS_EXT 0x34
#define CMD_READ_MULTIPLE 0xC4
#define CMD_READ_MULTIPLE_EXT 0x29
#define CMD_WRITE_MULTIPLE 0xC5
#define CMD_WRITE_MULTIPLE_EXT 0x39
#define CMD_SET_MULTIPLE_MODE 0xC6
#define CMD_STANDBY_IMMEDIATE 0xE0
#define CMD_STANDBY 0xE2
#define CMD_IDENTIFY 0xEC
#define CMD_SLEEP 0xE6
#define CMD_SET_FEATURES 0xEF
#define CMD_SECURITY_FREEZE_LOCK 0xF5
#ifdef HAVE_ATA_DMA
#define CMD_READ_DMA 0xC8
#define CMD_READ_DMA_EXT 0x25
#define CMD_WRITE_DMA 0xCA
#define CMD_WRITE_DMA_EXT 0x35
#endif
#define READWRITE_TIMEOUT 5*HZ
#ifdef HAVE_ATA_POWER_OFF
#define ATA_POWER_OFF_TIMEOUT 2*HZ
#endif
#if defined(HAVE_USBSTACK)
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
#define ATA_ACTIVE_IN_USB 1
#else
#define ATA_ACTIVE_IN_USB 0
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
enum {
ATA_BOOT = -1,
ATA_OFF,
ATA_SLEEPING,
ATA_SPINUP,
ATA_ON,
};
static int ata_state = ATA_BOOT;
static struct mutex ata_mtx SHAREDBSS_ATTR;
static int ata_device; /* device 0 (master) or 1 (slave) */
static int spinup_time = 0;
#if (CONFIG_LED == LED_REAL)
static bool ata_led_enabled = true;
static bool ata_led_on = false;
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
static long sleep_timeout = 5*HZ;
#ifdef HAVE_LBA48
static bool lba48 = false; /* set for 48 bit addressing */
#endif
static long last_disk_activity = -1;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
#ifdef HAVE_ATA_POWER_OFF
static long power_off_tick;
#endif
static unsigned long total_sectors;
static int multisectors; /* number of supported multisectors */
static unsigned short identify_info[SECTOR_SIZE/2];
#ifdef MAX_PHYS_SECTOR_SIZE
struct sector_cache_entry {
bool inuse;
unsigned long sectornum; /* logical sector */
unsigned char data[MAX_PHYS_SECTOR_SIZE];
};
/* buffer for reading and writing large physical sectors */
#define NUMCACHES 2
static struct sector_cache_entry sector_cache;
static int phys_sector_mult = 1;
#endif
#ifdef HAVE_ATA_DMA
static int dma_mode = 0;
#endif
#ifdef HAVE_ATA_POWER_OFF
static int ata_power_on(void);
#endif
static int perform_soft_reset(void);
static int set_multiple_mode(int sectors);
static int set_features(void);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
static inline void keep_ata_active(void)
{
last_disk_activity = current_tick;
}
static inline void schedule_ata_sleep(long from_now)
{
last_disk_activity = current_tick - sleep_timeout + from_now;
}
static inline bool ata_sleep_timed_out(void)
{
return sleep_timeout &&
TIME_AFTER(current_tick, last_disk_activity + sleep_timeout);
}
static inline void schedule_ata_power_off(void)
{
#ifdef HAVE_ATA_POWER_OFF
power_off_tick = current_tick + ATA_POWER_OFF_TIMEOUT;
#endif
}
static inline bool ata_power_off_timed_out(void)
{
#ifdef HAVE_ATA_POWER_OFF
return TIME_AFTER(current_tick, power_off_tick);
#else
return false;
#endif
}
#ifndef ATA_TARGET_POLLING
static ICODE_ATTR int wait_for_bsy(void)
{
long timeout = current_tick + HZ*30;
do
{
if (!(ATA_IN8(ATA_STATUS) & STATUS_BSY))
return 1;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
keep_ata_active();
yield();
} while (TIME_BEFORE(current_tick, timeout));
return 0; /* timeout */
}
static ICODE_ATTR int wait_for_rdy(void)
{
long timeout;
if (!wait_for_bsy())
return 0;
timeout = current_tick + HZ*10;
do
{
if (ATA_IN8(ATA_ALT_STATUS) & STATUS_RDY)
return 1;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
keep_ata_active();
yield();
} while (TIME_BEFORE(current_tick, timeout));
return 0; /* timeout */
}
#else
#define wait_for_bsy ata_wait_for_bsy
#define wait_for_rdy ata_wait_for_rdy
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
static int ata_perform_wakeup(int state)
{
if (state > ATA_OFF) {
if (perform_soft_reset()) {
return -1;
}
}
#ifdef HAVE_ATA_POWER_OFF
else {
if (ata_power_on()) {
return -2;
}
}
#endif
return 0;
}
static int ata_perform_sleep(void)
{
ATA_OUT8(ATA_SELECT, ata_device);
if(!wait_for_rdy()) {
DEBUGF("ata_perform_sleep() - not RDY\n");
return -1;
}
ATA_OUT8(ATA_COMMAND, CMD_SLEEP);
if (!wait_for_rdy())
{
DEBUGF("ata_perform_sleep() - CMD failed\n");
return -2;
}
return 0;
}
static ICODE_ATTR int wait_for_start_of_transfer(void)
{
if (!wait_for_bsy())
return 0;
return (ATA_IN8(ATA_ALT_STATUS) & (STATUS_BSY|STATUS_DRQ)) == STATUS_DRQ;
}
static ICODE_ATTR int wait_for_end_of_transfer(void)
{
if (!wait_for_bsy())
return 0;
return (ATA_IN8(ATA_ALT_STATUS) &
(STATUS_BSY|STATUS_RDY|STATUS_DF|STATUS_DRQ|STATUS_ERR))
== STATUS_RDY;
}
#if (CONFIG_LED == LED_REAL)
/* Conditionally block LED access for the ATA driver, so the LED can be
* (mis)used for other purposes */
static void ata_led(bool on)
{
ata_led_on = on;
if (ata_led_enabled)
led(ata_led_on);
}
#else
#define ata_led(on) led(on)
#endif
#ifndef ATA_OPTIMIZED_READING
static ICODE_ATTR void copy_read_sectors(unsigned char* buf, int wordcount)
{
unsigned short tmp = 0;
if ( (unsigned long)buf & 1)
{ /* not 16-bit aligned, copy byte by byte */
unsigned char* bufend = buf + wordcount*2;
do
{
tmp = ATA_IN16(ATA_DATA);
#if defined(ROCKBOX_LITTLE_ENDIAN)
*buf++ = tmp & 0xff; /* I assume big endian */
*buf++ = tmp >> 8; /* and don't use the SWAB16 macro */
#else
*buf++ = tmp >> 8;
*buf++ = tmp & 0xff;
#endif
} while (buf < bufend); /* tail loop is faster */
}
else
{ /* 16-bit aligned, can do faster copy */
unsigned short* wbuf = (unsigned short*)buf;
unsigned short* wbufend = wbuf + wordcount;
do
{
*wbuf = ATA_IN16(ATA_DATA);
} while (++wbuf < wbufend); /* tail loop is faster */
}
}
#endif /* !ATA_OPTIMIZED_READING */
#ifndef ATA_OPTIMIZED_WRITING
static ICODE_ATTR void copy_write_sectors(const unsigned char* buf,
int wordcount)
{
if ( (unsigned long)buf & 1)
{ /* not 16-bit aligned, copy byte by byte */
unsigned short tmp = 0;
const unsigned char* bufend = buf + wordcount*2;
do
{
#if defined(ROCKBOX_LITTLE_ENDIAN)
tmp = (unsigned short) *buf++;
tmp |= (unsigned short) *buf++ << 8;
#else
tmp = (unsigned short) *buf++ << 8;
tmp |= (unsigned short) *buf++;
#endif
ATA_OUT16(ATA_DATA, tmp);
} while (buf < bufend); /* tail loop is faster */
}
else
{ /* 16-bit aligned, can do faster copy */
unsigned short* wbuf = (unsigned short*)buf;
unsigned short* wbufend = wbuf + wordcount;
do
{
ATA_OUT16(ATA_DATA, *wbuf);
} while (++wbuf < wbufend); /* tail loop is faster */
}
}
#endif /* !ATA_OPTIMIZED_WRITING */
int ata_disk_isssd(void)
{
/* offset 217 is "Nominal Rotation rate"
0x0000 == Not reported
0x0001 == Solid State
0x0401 -> 0xffe == RPM
All others reserved
Some CF cards return 0x0100 (ie byteswapped 0x0001) so accept either
*/
return (identify_info[217] == 0x0001 || identify_info[217] == 0x0100);
}
static int ata_transfer_sectors(unsigned long start,
int incount,
void* inbuf,
int write)
{
int ret = 0;
long timeout;
int count;
void* buf;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
long spinup_start = spinup_start;
#ifdef HAVE_ATA_DMA
bool usedma = false;
#endif
if (start + incount > total_sectors) {
ret = -1;
goto error;
}
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
keep_ata_active();
ata_led(true);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state < ATA_ON) {
spinup_start = current_tick;
int state = ata_state;
ata_state = ATA_SPINUP;
if (ata_perform_wakeup(state)) {
ret = -2;
goto error;
}
}
timeout = current_tick + READWRITE_TIMEOUT;
ATA_OUT8(ATA_SELECT, ata_device);
if (!wait_for_rdy())
{
ret = -3;
goto error;
}
retry:
buf = inbuf;
count = incount;
while (TIME_BEFORE(current_tick, timeout)) {
ret = 0;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
keep_ata_active();
#ifdef HAVE_ATA_DMA
/* If DMA is supported and parameters are ok for DMA, use it */
if (dma_mode && ata_dma_setup(inbuf, incount * SECTOR_SIZE, write))
usedma = true;
#endif
#ifdef HAVE_LBA48
if (lba48)
{
ATA_OUT8(ATA_NSECTOR, count >> 8);
ATA_OUT8(ATA_NSECTOR, count & 0xff);
ATA_OUT8(ATA_SECTOR, (start >> 24) & 0xff); /* 31:24 */
ATA_OUT8(ATA_SECTOR, start & 0xff); /* 7:0 */
ATA_OUT8(ATA_LCYL, 0); /* 39:32 */
ATA_OUT8(ATA_LCYL, (start >> 8) & 0xff); /* 15:8 */
ATA_OUT8(ATA_HCYL, 0); /* 47:40 */
ATA_OUT8(ATA_HCYL, (start >> 16) & 0xff); /* 23:16 */
ATA_OUT8(ATA_SELECT, SELECT_LBA | ata_device);
#ifdef HAVE_ATA_DMA
if (write)
ATA_OUT8(ATA_COMMAND, usedma ? CMD_WRITE_DMA_EXT : CMD_WRITE_MULTIPLE_EXT);
else
ATA_OUT8(ATA_COMMAND, usedma ? CMD_READ_DMA_EXT : CMD_READ_MULTIPLE_EXT);
#else
ATA_OUT8(ATA_COMMAND, write ? CMD_WRITE_MULTIPLE_EXT : CMD_READ_MULTIPLE_EXT);
#endif
}
else
#endif
{
ATA_OUT8(ATA_NSECTOR, count & 0xff); /* 0 means 256 sectors */
ATA_OUT8(ATA_SECTOR, start & 0xff);
ATA_OUT8(ATA_LCYL, (start >> 8) & 0xff);
ATA_OUT8(ATA_HCYL, (start >> 16) & 0xff);
ATA_OUT8(ATA_SELECT, ((start >> 24) & 0xf) | SELECT_LBA | ata_device);
#ifdef HAVE_ATA_DMA
if (write)
ATA_OUT8(ATA_COMMAND, usedma ? CMD_WRITE_DMA : CMD_WRITE_MULTIPLE);
else
ATA_OUT8(ATA_COMMAND, usedma ? CMD_READ_DMA : CMD_READ_MULTIPLE);
#else
ATA_OUT8(ATA_COMMAND, write ? CMD_WRITE_MULTIPLE : CMD_READ_MULTIPLE);
#endif
}
/* wait at least 400ns between writing command and reading status */
__asm__ volatile ("nop");
__asm__ volatile ("nop");
__asm__ volatile ("nop");
__asm__ volatile ("nop");
__asm__ volatile ("nop");
#ifdef HAVE_ATA_DMA
if (usedma) {
if (!ata_dma_finish())
ret = -7;
if (ret != 0) {
perform_soft_reset();
goto retry;
}
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state == ATA_SPINUP) {
ata_state = ATA_ON;
spinup_time = current_tick - spinup_start;
}
}
else
#endif /* HAVE_ATA_DMA */
{
while (count) {
int sectors;
int wordcount;
int status;
int error;
if (!wait_for_start_of_transfer()) {
/* We have timed out waiting for RDY and/or DRQ, possibly
because the hard drive is shaking and has problems
reading the data. We have two options:
1) Wait some more
2) Perform a soft reset and try again.
We choose alternative 2.
*/
perform_soft_reset();
ret = -5;
goto retry;
}
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state == ATA_SPINUP) {
ata_state = ATA_ON;
spinup_time = current_tick - spinup_start;
}
/* read the status register exactly once per loop */
status = ATA_IN8(ATA_STATUS);
error = ATA_IN8(ATA_ERROR);
if (count >= multisectors)
sectors = multisectors;
else
sectors = count;
wordcount = sectors * SECTOR_SIZE / 2;
if (write)
copy_write_sectors(buf, wordcount);
else
copy_read_sectors(buf, wordcount);
/*
"Device errors encountered during READ MULTIPLE commands
are posted at the beginning of the block or partial block
transfer, but the DRQ bit is still set to one and the data
transfer shall take place, including transfer of corrupted
data, if any."
-- ATA specification
*/
if ( status & (STATUS_BSY | STATUS_ERR | STATUS_DF) ) {
perform_soft_reset();
ret = -6;
/* no point retrying IDNF, sector no. was invalid */
if (error & ERROR_IDNF)
break;
goto retry;
}
buf += sectors * SECTOR_SIZE; /* Advance one chunk of sectors */
count -= sectors;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
keep_ata_active();
}
}
if(!ret && !wait_for_end_of_transfer()) {
int error;
error = ATA_IN8(ATA_ERROR);
perform_soft_reset();
ret = -4;
/* no point retrying IDNF, sector no. was invalid */
if (error & ERROR_IDNF)
break;
goto retry;
}
break;
}
error:
ata_led(false);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ret < 0 && ata_state == ATA_SPINUP) {
/* bailed out before updating */
ata_state = ATA_ON;
}
return ret;
}
#ifndef MAX_PHYS_SECTOR_SIZE
int ata_read_sectors(IF_MD(int drive,)
unsigned long start,
int incount,
void* inbuf)
{
#ifdef HAVE_MULTIDRIVE
(void)drive; /* unused for now */
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
mutex_lock(&ata_mtx);
int rc = ata_transfer_sectors(start, incount, inbuf, false);
mutex_unlock(&ata_mtx);
return rc;
}
int ata_write_sectors(IF_MD(int drive,)
unsigned long start,
int count,
const void* buf)
{
#ifdef HAVE_MULTIDRIVE
(void)drive; /* unused for now */
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
mutex_lock(&ata_mtx);
int rc = ata_transfer_sectors(start, count, (void*)buf, true);
mutex_unlock(&ata_mtx);
return rc;
}
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
#endif /* ndef MAX_PHYS_SECTOR_SIZE */
#ifdef MAX_PHYS_SECTOR_SIZE
static int cache_sector(unsigned long sector)
{
int rc;
sector &= ~(phys_sector_mult - 1);
/* round down to physical sector boundary */
/* check whether the sector is already cached */
if (sector_cache.inuse && (sector_cache.sectornum == sector))
return 0;
/* not found: read the sector */
sector_cache.inuse = false;
rc = ata_transfer_sectors(sector, phys_sector_mult, sector_cache.data, false);
if (!rc)
{
sector_cache.sectornum = sector;
sector_cache.inuse = true;
}
return rc;
}
static inline int flush_current_sector(void)
{
return ata_transfer_sectors(sector_cache.sectornum, phys_sector_mult,
sector_cache.data, true);
}
int ata_read_sectors(IF_MD(int drive,)
unsigned long start,
int incount,
void* inbuf)
{
int rc = 0;
int offset;
#ifdef HAVE_MULTIDRIVE
(void)drive; /* unused for now */
#endif
mutex_lock(&ata_mtx);
offset = start & (phys_sector_mult - 1);
if (offset) /* first partial sector */
{
int partcount = MIN(incount, phys_sector_mult - offset);
rc = cache_sector(start);
if (rc)
{
rc = rc * 10 - 1;
goto error;
}
memcpy(inbuf, sector_cache.data + offset * SECTOR_SIZE,
partcount * SECTOR_SIZE);
start += partcount;
inbuf += partcount * SECTOR_SIZE;
incount -= partcount;
}
if (incount)
{
offset = incount & (phys_sector_mult - 1);
incount -= offset;
if (incount)
{
rc = ata_transfer_sectors(start, incount, inbuf, false);
if (rc)
{
rc = rc * 10 - 2;
goto error;
}
start += incount;
inbuf += incount * SECTOR_SIZE;
}
if (offset)
{
rc = cache_sector(start);
if (rc)
{
rc = rc * 10 - 3;
goto error;
}
memcpy(inbuf, sector_cache.data, offset * SECTOR_SIZE);
}
}
error:
mutex_unlock(&ata_mtx);
return rc;
}
int ata_write_sectors(IF_MD(int drive,)
unsigned long start,
int count,
const void* buf)
{
int rc = 0;
int offset;
#ifdef HAVE_MULTIDRIVE
(void)drive; /* unused for now */
#endif
mutex_lock(&ata_mtx);
offset = start & (phys_sector_mult - 1);
if (offset) /* first partial sector */
{
int partcount = MIN(count, phys_sector_mult - offset);
rc = cache_sector(start);
if (rc)
{
rc = rc * 10 - 1;
goto error;
}
memcpy(sector_cache.data + offset * SECTOR_SIZE, buf,
partcount * SECTOR_SIZE);
rc = flush_current_sector();
if (rc)
{
rc = rc * 10 - 2;
goto error;
}
start += partcount;
buf += partcount * SECTOR_SIZE;
count -= partcount;
}
if (count)
{
offset = count & (phys_sector_mult - 1);
count -= offset;
if (count)
{
rc = ata_transfer_sectors(start, count, (void*)buf, true);
if (rc)
{
rc = rc * 10 - 3;
goto error;
}
start += count;
buf += count * SECTOR_SIZE;
}
if (offset)
{
rc = cache_sector(start);
if (rc)
{
rc = rc * 10 - 4;
goto error;
}
memcpy(sector_cache.data, buf, offset * SECTOR_SIZE);
rc = flush_current_sector();
if (rc)
{
rc = rc * 10 - 5;
goto error;
}
}
}
error:
mutex_unlock(&ata_mtx);
return rc;
}
#endif /* MAX_PHYS_SECTOR_SIZE */
static int STORAGE_INIT_ATTR check_registers(void)
{
int i;
wait_for_bsy();
if (ATA_IN8(ATA_STATUS) & STATUS_BSY)
return -1;
for (i = 0; i<64; i++) {
ATA_OUT8(ATA_NSECTOR, TEST_PATTERN1);
ATA_OUT8(ATA_SECTOR, TEST_PATTERN2);
ATA_OUT8(ATA_LCYL, TEST_PATTERN3);
ATA_OUT8(ATA_HCYL, TEST_PATTERN4);
if (((ATA_IN8(ATA_NSECTOR) & 0xff) == TEST_PATTERN1) &&
((ATA_IN8(ATA_SECTOR) & 0xff) == TEST_PATTERN2) &&
((ATA_IN8(ATA_LCYL) & 0xff) == TEST_PATTERN3) &&
((ATA_IN8(ATA_HCYL) & 0xff) == TEST_PATTERN4))
return 0;
sleep(1);
}
return -2;
}
static int freeze_lock(void)
{
/* does the disk support Security Mode feature set? */
if (identify_info[82] & 2)
{
ATA_OUT8(ATA_SELECT, ata_device);
if (!wait_for_rdy())
return -1;
ATA_OUT8(ATA_COMMAND, CMD_SECURITY_FREEZE_LOCK);
if (!wait_for_rdy())
return -2;
}
return 0;
}
void ata_spindown(int seconds)
{
sleep_timeout = seconds * HZ;
}
bool ata_disk_is_active(void)
{
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
return ata_state >= ATA_SPINUP;
}
void ata_sleepnow(void)
{
/* Don't enter sleep if the device doesn't support
power management. */
if (!(identify_info[82] & (1 << 3)))
return;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state >= ATA_SPINUP) {
mutex_lock(&ata_mtx);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state == ATA_ON) {
if (!ata_perform_sleep()) {
ata_state = ATA_SLEEPING;
schedule_ata_power_off();
}
}
mutex_unlock(&ata_mtx);
}
}
void ata_spin(void)
{
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
keep_ata_active();
}
/* Hardware reset protocol as specified in chapter 9.1, ATA spec draft v5 */
#ifdef HAVE_ATA_POWER_OFF
static int ata_hard_reset(void)
#else
static int STORAGE_INIT_ATTR ata_hard_reset(void)
#endif
{
int ret;
mutex_lock(&ata_mtx);
ata_reset();
/* state HRR2 */
ATA_OUT8(ATA_SELECT, ata_device); /* select the right device */
ret = wait_for_bsy();
/* Massage the return code so it is 0 on success and -1 on failure */
ret = ret?0:-1;
mutex_unlock(&ata_mtx);
return ret;
}
// not putting this into STORAGE_INIT_ATTR, as ATA spec recommends to
// re-read identify_info after soft reset. So we'll do that.
static int identify(void)
{
int i;
ATA_OUT8(ATA_SELECT, ata_device);
if(!wait_for_rdy()) {
DEBUGF("identify() - not RDY\n");
return -1;
}
ATA_OUT8(ATA_COMMAND, CMD_IDENTIFY);
if (!wait_for_start_of_transfer())
{
DEBUGF("identify() - CMD failed\n");
return -2;
}
for (i=0; i<SECTOR_SIZE/2; i++) {
/* the IDENTIFY words are already swapped, so we need to treat
this info differently that normal sector data */
identify_info[i] = ATA_SWAP_IDENTIFY(ATA_IN16(ATA_DATA));
}
return 0;
}
static int perform_soft_reset(void)
{
/* If this code is allowed to run on a Nano, the next reads from the flash will
* time out, so we disable it. It shouldn't be necessary anyway, since the
* ATA -> Flash interface automatically sleeps almost immediately after the
* last command.
*/
int ret;
int retry_count;
ATA_OUT8(ATA_SELECT, SELECT_LBA | ata_device );
ATA_OUT8(ATA_CONTROL, CONTROL_nIEN|CONTROL_SRST );
sleep(1); /* >= 5us */
#ifdef HAVE_ATA_DMA
/* DMA requires INTRQ be enabled */
ATA_OUT8(ATA_CONTROL, 0);
#else
ATA_OUT8(ATA_CONTROL, CONTROL_nIEN);
#endif
sleep(1); /* >2ms */
/* This little sucker can take up to 30 seconds */
retry_count = 8;
do
{
ret = wait_for_rdy();
} while(!ret && retry_count--);
if (!ret)
return -1;
if (identify())
return -5;
if (set_features())
return -2;
if (set_multiple_mode(multisectors))
return -3;
if (freeze_lock())
return -4;
return 0;
}
int ata_soft_reset(void)
{
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
int ret = -6;
mutex_lock(&ata_mtx);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state > ATA_OFF) {
ret = perform_soft_reset();
}
mutex_unlock(&ata_mtx);
return ret;
}
#ifdef HAVE_ATA_POWER_OFF
static int ata_power_on(void)
{
int rc;
ide_power_enable(true);
sleep(HZ/4); /* allow voltage to build up */
/* Accessing the PP IDE controller too early after powering up the disk
* makes the core hang for a short time, causing an audio dropout. This
* also depends on the disk; iPod Mini G2 needs at least HZ/5 to get rid
* of the dropout. Since this time isn't additive (the wait_for_bsy() in
* ata_hard_reset() will shortened by the same amount), it's a good idea
* to do this on all HDD based targets. */
if( ata_hard_reset() )
return -1;
if (identify())
return -5;
rc = set_features();
if (rc)
return rc * 10 - 2;
if (set_multiple_mode(multisectors))
return -3;
if (freeze_lock())
return -4;
return 0;
}
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
#endif /* HAVE_ATA_POWER_OFF */
static int STORAGE_INIT_ATTR master_slave_detect(void)
{
/* master? */
ATA_OUT8(ATA_SELECT, 0);
if (ATA_IN8(ATA_STATUS) & (STATUS_RDY|STATUS_BSY)) {
ata_device = 0;
DEBUGF("Found master harddisk\n");
}
else {
/* slave? */
ATA_OUT8(ATA_SELECT, SELECT_DEVICE1);
if (ATA_IN8(ATA_STATUS) & (STATUS_RDY|STATUS_BSY)) {
ata_device = SELECT_DEVICE1;
DEBUGF("Found slave harddisk\n");
}
else
return -1;
}
return 0;
}
static int set_multiple_mode(int sectors)
{
ATA_OUT8(ATA_SELECT, ata_device);
if(!wait_for_rdy()) {
DEBUGF("set_multiple_mode() - not RDY\n");
return -1;
}
ATA_OUT8(ATA_NSECTOR, sectors);
ATA_OUT8(ATA_COMMAND, CMD_SET_MULTIPLE_MODE);
if (!wait_for_rdy())
{
DEBUGF("set_multiple_mode() - CMD failed\n");
return -2;
}
return 0;
}
#ifdef HAVE_ATA_DMA
static int get_best_mode(unsigned short identword, int max, int modetype)
{
unsigned short testbit = BIT_N(max);
while (1) {
if (identword & testbit)
return max | modetype;
testbit >>= 1;
if (!testbit)
return 0;
max--;
}
}
#endif
static int set_features(void)
{
static struct {
unsigned char id_word;
unsigned char id_bit;
unsigned char subcommand;
unsigned char parameter;
} features[] = {
{ 83, 14, 0x03, 0 }, /* force PIO mode */
{ 83, 3, 0x05, 0x80 }, /* adv. power management: lowest w/o standby */
{ 83, 9, 0x42, 0x80 }, /* acoustic management: lowest noise */
{ 82, 6, 0xaa, 0 }, /* enable read look-ahead */
#ifdef HAVE_ATA_DMA
{ 0, 0, 0x03, 0 }, /* DMA mode */
#endif
};
int i;
int pio_mode = 2;
/* Find out the highest supported PIO mode */
if (identify_info[53] & (1<<1)) { /* Is word 64 valid? */
if (identify_info[64] & 2)
pio_mode = 4;
else if(identify_info[64] & 1)
pio_mode = 3;
}
/* Update the table: set highest supported pio mode that we also support */
features[0].parameter = 8 + pio_mode;
#ifdef HAVE_ATA_DMA
if (identify_info[53] & (1<<2))
/* Ultra DMA mode info present, find a mode */
dma_mode = get_best_mode(identify_info[88], ATA_MAX_UDMA, 0x40);
if (!dma_mode) {
/* No UDMA mode found, try to find a multi-word DMA mode */
dma_mode = get_best_mode(identify_info[63], ATA_MAX_MWDMA, 0x20);
features[4].id_word = 63;
}
else
features[4].id_word = 88;
features[4].id_bit = dma_mode & 7;
features[4].parameter = dma_mode;
#endif /* HAVE_ATA_DMA */
ATA_OUT8(ATA_SELECT, ata_device);
if (!wait_for_rdy()) {
DEBUGF("set_features() - not RDY\n");
return -1;
}
for (i=0; i < (int)(sizeof(features)/sizeof(features[0])); i++) {
if (identify_info[features[i].id_word] & BIT_N(features[i].id_bit)) {
ATA_OUT8(ATA_FEATURE, features[i].subcommand);
ATA_OUT8(ATA_NSECTOR, features[i].parameter);
ATA_OUT8(ATA_COMMAND, CMD_SET_FEATURES);
if (!wait_for_rdy()) {
DEBUGF("set_features() - CMD failed\n");
return -10 - i;
}
if((ATA_IN8(ATA_ALT_STATUS) & STATUS_ERR) && (i != 1)) {
/* some CF cards don't like advanced powermanagement
even if they mark it as supported - go figure... */
if(ATA_IN8(ATA_ERROR) & ERROR_ABRT) {
return -20 - i;
}
}
}
}
#ifdef ATA_SET_PIO_TIMING
ata_set_pio_timings(pio_mode);
#endif
#ifdef HAVE_ATA_DMA
ata_dma_set_mode(dma_mode);
#endif
return 0;
}
unsigned short* ata_get_identify(void)
{
return identify_info;
}
static int STORAGE_INIT_ATTR init_and_check(bool hard_reset)
{
int rc;
if (hard_reset)
{
/* This should reset both master and slave, we don't yet know what's in */
ata_device = 0;
if (ata_hard_reset())
return -1;
}
rc = master_slave_detect();
if (rc)
return -10 + rc;
/* symptom fix: else check_registers() below may fail */
if (hard_reset && !wait_for_bsy())
return -20;
rc = check_registers();
if (rc)
return -30 + rc;
return 0;
}
int STORAGE_INIT_ATTR ata_init(void)
{
int rc = 0;
bool coldstart;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state == ATA_BOOT) {
mutex_init(&ata_mtx);
}
mutex_lock(&ata_mtx);
/* must be called before ata_device_init() */
coldstart = ata_is_coldstart();
ata_led(false);
ata_device_init();
ata_enable(true);
#ifdef MAX_PHYS_SECTOR_SIZE
memset(&sector_cache, 0, sizeof(sector_cache));
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (ata_state == ATA_BOOT) {
ata_state = ATA_OFF;
if (!ide_powered()) /* somebody has switched it off */
{
ide_power_enable(true);
sleep(HZ/4); /* allow voltage to build up */
}
#ifdef HAVE_ATA_DMA
/* DMA requires INTRQ be enabled */
ATA_OUT8(ATA_CONTROL, 0);
#endif
/* first try, hard reset at cold start only */
rc = init_and_check(coldstart);
if (rc)
{ /* failed? -> second try, always with hard reset */
DEBUGF("ata: init failed, retrying...\n");
rc = init_and_check(true);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (rc) {
goto error;
}
}
rc = identify();
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (rc) {
rc = -40 + rc;
goto error;
}
multisectors = identify_info[47] & 0xff;
if (multisectors == 0) /* Invalid multisector info, try with 16 */
multisectors = 16;
DEBUGF("ata: %d sectors per ata request\n",multisectors);
total_sectors = identify_info[60] | (identify_info[61] << 16);
#ifdef HAVE_LBA48
if (identify_info[83] & 0x0400 /* 48 bit address support */
&& total_sectors == 0x0FFFFFFF) /* and disk size >= 128 GiB */
{ /* (needs BigLBA addressing) */
if (identify_info[102] || identify_info[103])
panicf("Unsupported disk size: >= 2^32 sectors");
total_sectors = identify_info[100] | (identify_info[101] << 16);
lba48 = true; /* use BigLBA */
}
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
#endif /* HAVE_LBA48 */
rc = freeze_lock();
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (rc) {
rc = -50 + rc;
goto error;
}
rc = set_features();
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
if (rc) {
rc = -60 + rc;
goto error;
}
#ifdef MAX_PHYS_SECTOR_SIZE
/* Find out the physical sector size */
if((identify_info[106] & 0xe000) == 0x6000)
phys_sector_mult = BIT_N(identify_info[106] & 0x000f);
else
phys_sector_mult = 1;
DEBUGF("ata: %d logical sectors per phys sector", phys_sector_mult);
if (phys_sector_mult > 1)
{
/* Check if drive really needs emulation - if we can access
* sector 1 then assume the drive will handle it better than
* us, and ignore the large physical sectors.
*/
char throwaway[SECTOR_SIZE];
rc = ata_transfer_sectors(1, 1, &throwaway, false);
if (rc == 0)
phys_sector_mult = 1;
}
if (phys_sector_mult > (MAX_PHYS_SECTOR_SIZE/SECTOR_SIZE))
panicf("Unsupported physical sector size: %d",
phys_sector_mult * SECTOR_SIZE);
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
#endif /* MAX_PHYS_SECTOR_SIZE */
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
ata_state = ATA_ON;
keep_ata_active();
}
rc = set_multiple_mode(multisectors);
if (rc)
rc = -70 + rc;
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
error:
mutex_unlock(&ata_mtx);
return rc;
}
#if (CONFIG_LED == LED_REAL)
void ata_set_led_enabled(bool enabled)
{
ata_led_enabled = enabled;
if (ata_led_enabled)
led(ata_led_on);
else
led(false);
}
#endif
long ata_last_disk_activity(void)
{
return last_disk_activity;
}
int ata_spinup_time(void)
{
return spinup_time;
}
#ifdef STORAGE_GET_INFO
void ata_get_info(IF_MD(int drive,)struct storage_info *info)
{
unsigned short *src,*dest;
static char vendor[8];
static char product[16];
static char revision[4];
#ifdef HAVE_MULTIDRIVE
(void)drive; /* unused for now */
#endif
int i;
info->sector_size = SECTOR_SIZE;
info->num_sectors = total_sectors;
src = (unsigned short*)&identify_info[27];
dest = (unsigned short*)vendor;
for (i=0;i<4;i++)
dest[i] = htobe16(src[i]);
info->vendor=vendor;
src = (unsigned short*)&identify_info[31];
dest = (unsigned short*)product;
for (i=0;i<8;i++)
dest[i] = htobe16(src[i]);
info->product=product;
src = (unsigned short*)&identify_info[23];
dest = (unsigned short*)revision;
for (i=0;i<2;i++)
dest[i] = htobe16(src[i]);
info->revision=revision;
}
#endif
#ifdef HAVE_ATA_DMA
/* Returns last DMA mode as set by set_features() */
int ata_get_dma_mode(void)
{
return dma_mode;
}
/* Needed to allow updating while waiting for DMA to complete */
void ata_keep_active(void)
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
__attribute__((alias("ata_spin")));
#endif
#ifdef CONFIG_STORAGE_MULTI
int ata_num_drives(int first_drive)
{
/* We don't care which logical drive number(s) we have been assigned */
(void)first_drive;
return 1;
}
#endif
Unify storage threads into one * Editing a bunch of drivers' thread routines in order to implement a new feature is tedious. * No matter the number of storage drivers, they share one thread. No extra threads needed for CONFIG_STORAGE_MULTI. * Each has an event callback called by the storage thread. * A default callback is provided to fake sleeping in order to trigger idle callbacks. It could also do other default processing. Changes to it will be part of driver code without editing each one. * Drivers may sleep and wake as they please as long as they give a low pulse on their storage bit to ask to go into sleep mode. Idle callback is called on its behalf and driver immediately put into sleep mode. * Drivers may indicate they are to continue receiving events in USB mode, otherwise they receve nothing until disconnect (they do receive SYS_USB_DISCONNECTED no matter what). * Rework a few things to keep the callback implementation sane and maintainable. ata.c was dreadful with all those bools; make it a state machine and easier to follow. Remove last_user_activity; it has no purpose that isn't served by keeping the disk active through last_disk_activity instead. * Even-out stack sizes partly because of a lack of a decent place to define them by driver or SoC or whatever; it doesn't seem too critical to do that anyway. Many are simply too large while at least one isn't really adequate. They may be individually overridden if necessary (figure out where). The thread uses the greatest size demanded. Newer file code is much more frugal with stack space. I barely see use crack 50% after idle callbacks (usually mid-40s). Card insert/eject doesn't demand much. * No forcing of idle callbacks. If it isn't necessary for one or more non-disk storage types, it really isn't any more necessary for disk storage. Besides, it makes the whole thing easier to implement. Change-Id: Id30c284d82a8af66e47f2cfe104c52cbd8aa7215
2017-03-15 05:51:54 +00:00
int ata_event(long id, intptr_t data)
{
int rc = 0;
/* GCC does a lousy job culling unreachable cases in the default handler
if statements are in a switch statement, so we'll do it this way. Only
the first case is frequently hit anyway. */
if (LIKELY(id == Q_STORAGE_TICK)) {
/* won't see ATA_BOOT in here */
int state = ata_state;
if (state != ATA_ON || !ata_sleep_timed_out()) {
if (state == ATA_SLEEPING && ata_power_off_timed_out()) {
mutex_lock(&ata_mtx);
if (ata_state == ATA_SLEEPING) {
ide_power_enable(false);
ata_state = ATA_OFF;
}
mutex_unlock(&ata_mtx);
}
STG_EVENT_ASSERT_ACTIVE(STORAGE_ATA);
}
}
else if (id == Q_STORAGE_SLEEPNOW) {
ata_sleepnow();
}
else if (id == Q_STORAGE_SLEEP) {
schedule_ata_sleep(HZ/5);
}
#ifndef USB_NONE
else if (id == SYS_USB_CONNECTED) {
if (ATA_ACTIVE_IN_USB) {
/* There is no need to force ATA power on */
STG_EVENT_ASSERT_ACTIVE(STORAGE_ATA);
}
else {
mutex_lock(&ata_mtx);
if (ata_state < ATA_ON) {
ata_led(true);
if (!(rc = ata_perform_wakeup(ata_state))) {
ata_state = ATA_ON;
}
ata_led(false);
}
mutex_unlock(&ata_mtx);
}
}
#endif /* ndef USB_NONE */
else {
rc = storage_event_default_handler(id, data, last_disk_activity,
STORAGE_ATA);
}
return rc;
}