rockbox/firmware/target/arm/pp/ata-sd-pp.c

1294 lines
37 KiB
C
Raw Normal View History

/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
* Copyright (C) 2006 Daniel Ankers
*
* 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 "config.h" /* for HAVE_MULTIDRIVE */
#include "sdmmc.h"
#include "gcc_extensions.h"
#ifdef HAVE_HOTSWAP
#include "sd-pp-target.h"
#endif
#include "system.h"
#include <string.h>
#include "led.h"
#include "cpu.h"
#include "storage.h"
#include "fs_defines.h"
#define BLOCKS_PER_BANK 0x7a7800
/* Comparing documentations of various MMC/SD controllers revealed, */
/* that this controller seems to be a mix of PXA27x, PXA255 and */
/* some PP specific stuff. The register and bit definitions are */
/* taken from the 'PXA27x Developers Manual', as it appears to be */
/* the closest match. Known differences and obscurities are commented.*/
#define MMC_STRPCL (*(volatile unsigned int *)(0x70008200))
#define MMC_STAT (*(volatile unsigned int *)(0x70008204))
#define MMC_CLKRT (*(volatile unsigned int *)(0x70008208))
#define MMC_SPI (*(volatile unsigned int *)(0x7000820c))
#define MMC_CMDAT (*(volatile unsigned int *)(0x70008210))
#define MMC_RESTO (*(volatile unsigned int *)(0x70008214))
#define MMC_RDTO (*(volatile unsigned int *)(0x70008218))
#define MMC_BLKLEN (*(volatile unsigned int *)(0x7000821c))
#define MMC_NUMBLK (*(volatile unsigned int *)(0x70008220))
#define MMC_I_MASK (*(volatile unsigned int *)(0x70008224))
#define MMC_CMD (*(volatile unsigned int *)(0x70008228))
#define MMC_ARGH (*(volatile unsigned int *)(0x7000822c))
#define MMC_ARGL (*(volatile unsigned int *)(0x70008230))
#define MMC_RES (*(volatile unsigned int *)(0x70008234))
/* PXA255/27x have separate RX/TX FIFOs with 32x8 bit */
/* PP502x has a combined Data FIFO with 16x16 bit */
#define MMC_DATA_FIFO (*(volatile unsigned int *)(0x70008280))
/* PP specific registers, no other controller seem to have such. */
#define MMC_SD_STATE (*(volatile unsigned int *)(0x70008238))
#define MMC_INIT_1 (*(volatile unsigned int *)(0x70008240))
#define MMC_INIT_2 (*(volatile unsigned int *)(0x70008244))
/* MMC_STAT bits */
#define STAT_SDIO_SUSPEND_ACK (1 << 16)
#define STAT_SDIO_INT (1 << 15)
#define STAT_RD_STALLED (1 << 14)
#define STAT_END_CMD_RES (1 << 13)
#define STAT_PRG_DONE (1 << 12)
#define STAT_DATA_TRAN_DONE (1 << 11)
#define STAT_SPI_WR_ERR (1 << 10)
#define STAT_FLASH_ERR (1 << 9)
#define STAT_CLK_EN (1 << 8)
#define STAT_RECV_FIFO_FULL (1 << 7) /* taken from PXA255 */
#define STAT_XMIT_FIFO_EMPTY (1 << 6) /* taken from PXA255 */
#define STAT_RES_CRC_ERR (1 << 5)
#define STAT_DAT_ERR_TOKEN (1 << 4)
#define STAT_CRC_RD_ERR (1 << 3)
#define STAT_CRC_WR_ERR (1 << 2)
#define STAT_TIME_OUT_RES (1 << 1)
#define STAT_TIME_OUT_READ (1)
#define STAT_ERROR_BITS (0x3f)
/* MMC_CMDAT bits */
/* Some of the bits used by the OF don't make much sense with these */
/* definitions. So they're probably different between PXA and PP502x */
/* Bits 0-5 appear to match though. */
#define CMDAT_SDIO_RESUME (1 << 13)
#define CMDAT_SDIO_SUSPEND (1 << 12)
#define CMDAT_SDIO_INT_EN (1 << 11)
#define CMDAT_STOP_TRAN (1 << 10)
#define CMDAT_SD_4DAT (1 << 8)
#define CMDAT_DMA_EN (1 << 7)
#define CMDAT_INIT (1 << 6)
#define CMDAT_BUSY (1 << 5)
#define CMDAT_STRM_BLK (1 << 4)
#define CMDAT_WR_RD (1 << 3)
#define CMDAT_DATA_EN (1 << 2)
#define CMDAT_RES_TYPE3 (3)
#define CMDAT_RES_TYPE2 (2)
#define CMDAT_RES_TYPE1 (1)
/* MMC_I_MASK bits */
/* PP502x apparently only has bits 0-3 */
#define I_MASK_SDIO_SUSPEND_ACK (1 << 12)
#define I_MASK_SDIO_INT (1 << 11)
#define I_MASK_RD_STALLED (1 << 10)
#define I_MASK_RES_ERR (1 << 9)
#define I_MASK_DAT_ERR (1 << 8)
#define I_MASK_TINT (1 << 7)
#define I_MASK_TXFIFO_WR_REQ (1 << 6)
#define I_MASK_RXFIFO_RD_REQ (1 << 5)
#define I_MASK_CLK_IS_OFF (1 << 4)
#define I_MASK_STOP_CMD (1 << 3)
#define I_MASK_END_CMD_RES (1 << 2)
#define I_MASK_PRG_DONE (1 << 1)
#define I_MASK_DATA_TRAN_DONE (1 << 0)
#define FIFO_LEN 16 /* FIFO is 16 words deep */
#define EC_OK 0
#define EC_FAILED 1
#define EC_NOCARD 2
#define EC_WAIT_STATE_FAILED 3
#define EC_CHECK_TIMEOUT_FAILED 4
#define EC_POWER_UP 5
#define EC_READ_TIMEOUT 6
#define EC_WRITE_TIMEOUT 7
#define EC_TRAN_SEL_BANK 8
#define EC_TRAN_READ_ENTRY 9
#define EC_TRAN_READ_EXIT 10
#define EC_TRAN_WRITE_ENTRY 11
#define EC_TRAN_WRITE_EXIT 12
#define EC_FIFO_SEL_BANK_EMPTY 13
#define EC_FIFO_SEL_BANK_DONE 14
#define EC_FIFO_ENA_BANK_EMPTY 15
#define EC_FIFO_READ_FULL 16
#define EC_FIFO_WR_EMPTY 17
#define EC_FIFO_WR_DONE 18
#define EC_COMMAND 19
#define NUM_EC 20
/* for compatibility */
static long last_disk_activity = -1;
static long next_yield = 0;
#define MIN_YIELD_PERIOD 1000
static tCardInfo card_info[2];
static tCardInfo *currcard = NULL; /* current active card */
struct sd_card_status
{
int retry;
int retry_max;
};
static struct sd_card_status sd_status[NUM_DRIVES] =
{
{ 0, 1 },
#ifdef HAVE_MULTIDRIVE
{ 0, 10 }
#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 struct mutex sd_mtx SHAREDBSS_ATTR;
#ifdef HAVE_HOTSWAP
static int sd_first_drive = 0;
#endif
/* Private Functions */
static unsigned int check_time[NUM_EC];
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 enable_controller(bool on)
{
if(on)
{
DEV_EN |= DEV_ATA; /* Enable controller */
}
else
{
DEV_EN &= ~DEV_ATA; /* Disable controller */
}
}
static inline bool sd_check_timeout(long timeout, int id)
{
return !TIME_AFTER(USEC_TIMER, check_time[id] + timeout);
}
static bool sd_poll_status(unsigned int trigger, long timeout)
{
long t = USEC_TIMER;
while ((MMC_STAT & trigger) == 0)
{
long time = USEC_TIMER;
if (TIME_AFTER(time, next_yield))
{
long ty = USEC_TIMER;
yield();
timeout += USEC_TIMER - ty;
next_yield = ty + MIN_YIELD_PERIOD;
}
if (TIME_AFTER(time, t + timeout))
return false;
}
return true;
}
static int sd_command(unsigned int cmd, unsigned long arg1,
unsigned long *response, unsigned int cmdat)
{
int i, words; /* Number of 16 bit words to read from MMC_RES */
unsigned int data[9];
MMC_CMD = cmd;
MMC_ARGH = (unsigned int)((arg1 & 0xffff0000) >> 16);
MMC_ARGL = (unsigned int)((arg1 & 0xffff));
MMC_CMDAT = cmdat;
if (!sd_poll_status(STAT_END_CMD_RES, 100000))
return -EC_COMMAND;
if ((MMC_STAT & STAT_ERROR_BITS) != 0)
/* Error sending command */
return -EC_COMMAND - (MMC_STAT & STAT_ERROR_BITS)*100;
if (cmd == SD_GO_IDLE_STATE)
return 0; /* no response here */
words = (cmdat == CMDAT_RES_TYPE2) ? 9 : 3;
for (i = 0; i < words; i++) /* MMC_RES is read MSB first */
data[i] = MMC_RES; /* Read most significant 16-bit word */
if (response == NULL)
{
/* response discarded */
}
else if (cmdat == CMDAT_RES_TYPE2)
{
/* Response type 2 has the following structure:
* [135:135] Start Bit - '0'
* [134:134] Transmission bit - '0'
* [133:128] Reserved - '111111'
* [127:001] CID or CSD register including internal CRC7
* [000:000] End Bit - '1'
*/
response[3] = (data[0]<<24) + (data[1]<<8) + (data[2]>>8);
response[2] = (data[2]<<24) + (data[3]<<8) + (data[4]>>8);
response[1] = (data[4]<<24) + (data[5]<<8) + (data[6]>>8);
response[0] = (data[6]<<24) + (data[7]<<8) + (data[8]>>8);
}
else
{
/* Response types 1, 1b, 3, 6, 7 have the following structure:
* Types 4 and 5 are not supported.
*
* [47] Start bit - '0'
* [46] Transmission bit - '0'
* [45:40] R1, R1b, R6, R7: Command index
* R3: Reserved - '111111'
* [39:8] R1, R1b: Card Status
* R3: OCR Register
* R6: [31:16] RCA
* [15: 0] Card Status Bits 23, 22, 19, 12:0
* [23] COM_CRC_ERROR
* [22] ILLEGAL_COMMAND
* [19] ERROR
* [12:9] CURRENT_STATE
* [8] READY_FOR_DATA
* [7:6]
* [5] SD_APP_CMD
* [4]
* [3] AKE_SEQ_ERROR
* [2] Reserved
* [1:0] Reserved for test mode
* R7: [19:16] Voltage accepted
* [15:8] echo-back of check pattern
* [7:1] R1, R1b: CRC7
* R3: Reserved - '1111111'
* [0] End Bit - '1'
*/
response[0] = (data[0]<<24) + (data[1]<<8) + (data[2]>>8);
}
return 0;
}
static int sd_wait_for_state(unsigned int state, int id)
{
unsigned long response = 0;
unsigned int timeout = 0x80000;
check_time[id] = USEC_TIMER;
while (1)
{
int ret = sd_command(SD_SEND_STATUS, currcard->rca, &response, CMDAT_RES_TYPE1);
long us;
if (ret < 0)
return ret*100 - id;
if (((response >> 9) & 0xf) == state)
{
MMC_SD_STATE = state;
return 0;
}
if (!sd_check_timeout(timeout, id))
return -EC_WAIT_STATE_FAILED*100 - id;
us = USEC_TIMER;
if (TIME_AFTER(us, next_yield))
{
yield();
timeout += USEC_TIMER - us;
next_yield = us + MIN_YIELD_PERIOD;
}
}
}
static inline bool card_detect_target(void)
{
#ifdef HAVE_HOTSWAP
#ifdef SANSA_E200
return (GPIOA_INPUT_VAL & 0x80) == 0; /* low active */
#elif defined SANSA_C200
return (GPIOL_INPUT_VAL & 0x08) != 0; /* high active */
#endif
#else
return false;
#endif
}
static inline void copy_read_sectors_fast(unsigned char **buf)
{
/* Copy one chunk of 16 words using best method for start alignment */
switch ( (intptr_t)*buf & 3 )
{
case 0:
asm volatile (
"ldmia %[data], { r2-r9 } \r\n"
"orr r2, r2, r3, lsl #16 \r\n"
"orr r4, r4, r5, lsl #16 \r\n"
"orr r6, r6, r7, lsl #16 \r\n"
"orr r8, r8, r9, lsl #16 \r\n"
"stmia %[buf]!, { r2, r4, r6, r8 } \r\n"
"ldmia %[data], { r2-r9 } \r\n"
"orr r2, r2, r3, lsl #16 \r\n"
"orr r4, r4, r5, lsl #16 \r\n"
"orr r6, r6, r7, lsl #16 \r\n"
"orr r8, r8, r9, lsl #16 \r\n"
"stmia %[buf]!, { r2, r4, r6, r8 } \r\n"
: [buf]"+&r"(*buf)
: [data]"r"(&MMC_DATA_FIFO)
: "r2", "r3", "r4", "r5", "r6", "r7", "r8", "r9"
);
break;
case 1:
asm volatile (
"ldmia %[data], { r2-r9 } \r\n"
"orr r3, r2, r3, lsl #16 \r\n"
"strb r3, [%[buf]], #1 \r\n"
"mov r3, r3, lsr #8 \r\n"
"strh r3, [%[buf]], #2 \r\n"
"mov r3, r3, lsr #16 \r\n"
"orr r3, r3, r4, lsl #8 \r\n"
"orr r3, r3, r5, lsl #24 \r\n"
"mov r5, r5, lsr #8 \r\n"
"orr r5, r5, r6, lsl #8 \r\n"
"orr r5, r5, r7, lsl #24 \r\n"
"mov r7, r7, lsr #8 \r\n"
"orr r7, r7, r8, lsl #8 \r\n"
"orr r7, r7, r9, lsl #24 \r\n"
"mov r2, r9, lsr #8 \r\n"
"stmia %[buf]!, { r3, r5, r7 } \r\n"
"ldmia %[data], { r3-r10 } \r\n"
"orr r2, r2, r3, lsl #8 \r\n"
"orr r2, r2, r4, lsl #24 \r\n"
"mov r4, r4, lsr #8 \r\n"
"orr r4, r4, r5, lsl #8 \r\n"
"orr r4, r4, r6, lsl #24 \r\n"
"mov r6, r6, lsr #8 \r\n"
"orr r6, r6, r7, lsl #8 \r\n"
"orr r6, r6, r8, lsl #24 \r\n"
"mov r8, r8, lsr #8 \r\n"
"orr r8, r8, r9, lsl #8 \r\n"
"orr r8, r8, r10, lsl #24 \r\n"
"mov r10, r10, lsr #8 \r\n"
"stmia %[buf]!, { r2, r4, r6, r8 } \r\n"
"strb r10, [%[buf]], #1 \r\n"
: [buf]"+&r"(*buf)
: [data]"r"(&MMC_DATA_FIFO)
: "r2", "r3", "r4", "r5", "r6", "r7", "r8", "r9", "r10"
);
break;
case 2:
asm volatile (
"ldmia %[data], { r2-r9 } \r\n"
"strh r2, [%[buf]], #2 \r\n"
"orr r3, r3, r4, lsl #16 \r\n"
"orr r5, r5, r6, lsl #16 \r\n"
"orr r7, r7, r8, lsl #16 \r\n"
"stmia %[buf]!, { r3, r5, r7 } \r\n"
"ldmia %[data], { r2-r8, r10 } \r\n"
"orr r2, r9, r2, lsl #16 \r\n"
"orr r3, r3, r4, lsl #16 \r\n"
"orr r5, r5, r6, lsl #16 \r\n"
"orr r7, r7, r8, lsl #16 \r\n"
"stmia %[buf]!, { r2, r3, r5, r7 } \r\n"
"strh r10, [%[buf]], #2 \r\n"
: [buf]"+&r"(*buf)
: [data]"r"(&MMC_DATA_FIFO)
: "r2", "r3", "r4", "r5", "r6", "r7", "r8", "r9", "r10"
);
break;
case 3:
asm volatile (
"ldmia %[data], { r2-r9 } \r\n"
"orr r3, r2, r3, lsl #16 \r\n"
"strb r3, [%[buf]], #1 \r\n"
"mov r3, r3, lsr #8 \r\n"
"orr r3, r3, r4, lsl #24 \r\n"
"mov r4, r4, lsr #8 \r\n"
"orr r5, r4, r5, lsl #8 \r\n"
"orr r5, r5, r6, lsl #24 \r\n"
"mov r6, r6, lsr #8 \r\n"
"orr r7, r6, r7, lsl #8 \r\n"
"orr r7, r7, r8, lsl #24 \r\n"
"mov r8, r8, lsr #8 \r\n"
"orr r2, r8, r9, lsl #8 \r\n"
"stmia %[buf]!, { r3, r5, r7 } \r\n"
"ldmia %[data], { r3-r10 } \r\n"
"orr r2, r2, r3, lsl #24 \r\n"
"mov r3, r3, lsr #8 \r\n"
"orr r4, r3, r4, lsl #8 \r\n"
"orr r4, r4, r5, lsl #24 \r\n"
"mov r5, r5, lsr #8 \r\n"
"orr r6, r5, r6, lsl #8 \r\n"
"orr r6, r6, r7, lsl #24 \r\n"
"mov r7, r7, lsr #8 \r\n"
"orr r8, r7, r8, lsl #8 \r\n"
"orr r8, r8, r9, lsl #24 \r\n"
"mov r9, r9, lsr #8 \r\n"
"orr r10, r9, r10, lsl #8 \r\n"
"stmia %[buf]!, { r2, r4, r6, r8 } \r\n"
"strh r10, [%[buf]], #2 \r\n"
"mov r10, r10, lsr #16 \r\n"
"strb r10, [%[buf]], #1 \r\n"
: [buf]"+&r"(*buf)
: [data]"r"(&MMC_DATA_FIFO)
: "r2", "r3", "r4", "r5", "r6", "r7", "r8", "r9", "r10"
);
break;
}
}
static inline void copy_read_sectors_slow(unsigned char** buf)
{
int cnt = FIFO_LEN;
int t;
/* Copy one chunk of 16 words */
asm volatile (
"1: \r\n"
"ldrh %[t], [%[data]] \r\n"
"strb %[t], [%[buf]], #1 \r\n"
"mov %[t], %[t], lsr #8 \r\n"
"strb %[t], [%[buf]], #1 \r\n"
"subs %[cnt], %[cnt], #1 \r\n"
"bgt 1b \r\n"
: [cnt]"+&r"(cnt), [buf]"+&r"(*buf),
[t]"=&r"(t)
: [data]"r"(&MMC_DATA_FIFO)
);
}
/* Writes have to be kept slow for now */
static inline void copy_write_sectors(const unsigned char** buf)
{
int cnt = FIFO_LEN - 1;
unsigned t;
long time;
time = USEC_TIMER + 3;
if (((intptr_t)*buf & 3) == 0)
{
asm volatile (
"ldmia %[buf]!, { r3, r5, r7, r9 } \r\n"
"mov r4, r3, lsr #16 \r\n"
"mov r6, r5, lsr #16 \r\n"
"mov r8, r7, lsr #16 \r\n"
"mov r10, r9, lsr #16 \r\n"
"stmia %[data], { r3-r10 } \r\n"
"ldmia %[buf]!, { r3, r5, r7, r9 } \r\n"
"mov r4, r3, lsr #16 \r\n"
"mov r6, r5, lsr #16 \r\n"
"mov r8, r7, lsr #16 \r\n"
"mov %[t], r9, lsr #16 \r\n"
"stmia %[data], { r3-r9 } \r\n"
: [buf]"+&r"(*buf), [t]"=&r"(t)
: [data]"r"(&MMC_DATA_FIFO)
: "r3", "r4", "r5", "r6", "r7", "r8", "r9", "r10"
);
}
else
{
do
{
t = *(*buf)++;
t |= *(*buf)++ << 8;
MMC_DATA_FIFO = t;
} while (--cnt > 0); /* tail loop is faster */
t = *(*buf)++;
t |= *(*buf)++ << 8;
}
/* Don't write the last word before at least 3 usec have elapsed since FIFO_EMPTY */
/* This prevents the 'two bytes inserted' bug. */
while (!TIME_AFTER(USEC_TIMER, time));
MMC_DATA_FIFO = t;
}
static int sd_select_bank(unsigned char bank)
{
unsigned char card_data[FIFO_LEN*2];// FIFO_LEN words=FIFO_LEN*2 bytes
const unsigned char* write_buf;
int i, ret;
memset(card_data, 0, sizeof card_data);
ret = sd_wait_for_state(SD_TRAN, EC_TRAN_SEL_BANK);
if (ret < 0)
return ret;
MMC_BLKLEN = 512;
MMC_NUMBLK = 1;
ret = sd_command(35, 0, NULL, /* CMD35 is vendor specific */
0x1c00 | CMDAT_WR_RD | CMDAT_DATA_EN | CMDAT_RES_TYPE1);
if (ret < 0)
return ret;
MMC_SD_STATE = SD_PRG;
card_data[0] = bank;
/* Write the card data */
for (i = 0; i < SD_BLOCK_SIZE/2; i += FIFO_LEN)
{
write_buf = card_data;
/* Wait for the FIFO to empty */
if (sd_poll_status(STAT_XMIT_FIFO_EMPTY, 10000))
{
copy_write_sectors(&write_buf); /* Copy one chunk of 16 words */
/* clear buffer: only the first chunk contains interesting data (bank), the remaining is zero filling */
memset(card_data, 0, sizeof card_data);
continue;
}
return -EC_FIFO_SEL_BANK_EMPTY;
}
if (!sd_poll_status(STAT_PRG_DONE, 10000))
return -EC_FIFO_SEL_BANK_DONE;
currcard->current_bank = bank;
return 0;
}
static void sd_card_mux(int card_no)
{
/* Set the current card mux */
#if defined(SANSA_E200)
if (card_no == 0)
{
GPO32_VAL |= 0x4;
GPIO_CLEAR_BITWISE(GPIOA_ENABLE, 0x7a);
GPIO_CLEAR_BITWISE(GPIOA_OUTPUT_EN, 0x7a);
GPIO_SET_BITWISE(GPIOD_ENABLE, 0x1f);
GPIO_SET_BITWISE(GPIOD_OUTPUT_VAL, 0x1f);
GPIO_SET_BITWISE(GPIOD_OUTPUT_EN, 0x1f);
outl((inl(0x70000014) & ~(0x3ffff)) | 0x255aa, 0x70000014);
}
else
{
GPO32_VAL &= ~0x4;
GPIO_CLEAR_BITWISE(GPIOD_ENABLE, 0x1f);
GPIO_CLEAR_BITWISE(GPIOD_OUTPUT_EN, 0x1f);
GPIO_SET_BITWISE(GPIOA_ENABLE, 0x7a);
GPIO_SET_BITWISE(GPIOA_OUTPUT_VAL, 0x7a);
GPIO_SET_BITWISE( GPIOA_OUTPUT_EN, 0x7a);
outl(inl(0x70000014) & ~(0x3ffff), 0x70000014);
}
#elif defined(SANSA_C200)
if (card_no == 0)
{
GPO32_VAL |= 0x4;
GPIO_CLEAR_BITWISE(GPIOD_ENABLE, 0x1f);
GPIO_CLEAR_BITWISE(GPIOD_OUTPUT_EN, 0x1f);
GPIO_SET_BITWISE(GPIOA_ENABLE, 0x7a);
GPIO_SET_BITWISE(GPIOA_OUTPUT_VAL, 0x7a);
GPIO_SET_BITWISE( GPIOA_OUTPUT_EN, 0x7a);
outl(inl(0x70000014) & ~(0x3ffff), 0x70000014);
}
else
{
GPO32_VAL &= ~0x4;
GPIO_CLEAR_BITWISE(GPIOA_ENABLE, 0x7a);
GPIO_CLEAR_BITWISE(GPIOA_OUTPUT_EN, 0x7a);
GPIO_SET_BITWISE(GPIOD_ENABLE, 0x1f);
GPIO_SET_BITWISE(GPIOD_OUTPUT_VAL, 0x1f);
GPIO_SET_BITWISE(GPIOD_OUTPUT_EN, 0x1f);
outl((inl(0x70000014) & ~(0x3ffff)) | 0x255aa, 0x70000014);
}
#elif defined(PHILIPS_SA9200)
/* only 1 "card" (no external memory card) */
(void)card_no;
GPIO_SET_BITWISE(GPIOH_ENABLE, 0x80);
GPIO_SET_BITWISE(GPIOH_OUTPUT_EN, 0x80);
outl(0x255aa, 0x70000014);
GPIO_CLEAR_BITWISE(GPIOA_ENABLE, 0x04);
GPIO_CLEAR_BITWISE(GPIOA_OUTPUT_EN, 0x04);
GPIO_CLEAR_BITWISE(GPIOA_ENABLE, 0x7a);
GPIO_CLEAR_BITWISE(GPIOA_OUTPUT_EN, 0x7a);
GPIO_SET_BITWISE(GPIOH_OUTPUT_VAL, 0x80);
GPIO_SET_BITWISE(GPIOH_OUTPUT_EN, 0x80);
#endif
}
static void sd_init_device(int card_no)
{
/* SD Protocol registers */
#ifdef HAVE_HOTSWAP
unsigned long response = 0;
#endif
unsigned int i;
unsigned char carddata[512];
unsigned char *dataptr;
unsigned long temp_reg[4];
int ret;
/* Enable and initialise controller */
MMC_CLKRT = 6; /* switch to lowest clock rate */
/* Initialise card data as blank */
memset(currcard, 0, sizeof(*currcard));
/* Switch card mux to card to initialize */
sd_card_mux(card_no);
/* Init NAND */
MMC_INIT_1 |= (1 << 15);
MMC_INIT_2 |= (1 << 15);
MMC_INIT_2 &= ~(3 << 12);
MMC_INIT_2 |= (1 << 13);
MMC_INIT_1 &= ~(3 << 12);
MMC_INIT_1 |= (1 << 13);
DEV_EN |= DEV_ATA; /* Enable controller */
DEV_RS |= DEV_ATA; /* Reset controller */
DEV_RS &=~DEV_ATA; /* Clear Reset */
MMC_SD_STATE = SD_TRAN;
MMC_I_MASK = 0xf; /* disable interrupts */
ret = sd_command(SD_GO_IDLE_STATE, 0, NULL, 0x100);
if (ret < 0)
goto card_init_error;
check_time[EC_POWER_UP] = USEC_TIMER;
#ifdef HAVE_HOTSWAP
/* Check for SDHC:
- non-SDHC cards simply ignore SD_SEND_IF_COND (CMD8) and we get error -219,
which we can just ignore and assume we're dealing with standard SD.
- SDHC cards echo back the argument into the response. This is how we
tell if the card is SDHC.
*/
ret = sd_command(SD_SEND_IF_COND,0x1aa, &response,
CMDAT_DATA_EN | CMDAT_RES_TYPE3);
if ( (ret < 0) && (ret!=-219) )
goto card_init_error;
#endif
while ((currcard->ocr & (1 << 31)) == 0) /* until card is powered up */
{
ret = sd_command(SD_APP_CMD, currcard->rca, NULL, CMDAT_RES_TYPE1);
if (ret < 0)
goto card_init_error;
#ifdef HAVE_HOTSWAP
if(response == 0x1aa)
{
/* SDHC */
ret = sd_command(SD_APP_OP_COND, (1<<30)|0x100000,
&currcard->ocr, CMDAT_RES_TYPE3);
}
else
#endif /* HAVE_HOTSWAP */
{
/* SD Standard */
ret = sd_command(SD_APP_OP_COND, 0x100000, &currcard->ocr,
CMDAT_RES_TYPE3);
}
if (ret < 0)
goto card_init_error;
if (!sd_check_timeout(5000000, EC_POWER_UP))
{
ret = -EC_POWER_UP;
goto card_init_error;
}
}
ret = sd_command(SD_ALL_SEND_CID, 0, temp_reg, CMDAT_RES_TYPE2);
if (ret < 0)
goto card_init_error;
for(i=0; i<4; i++)
currcard->cid[i] = temp_reg[3-i];
ret = sd_command(SD_SEND_RELATIVE_ADDR, 0, &currcard->rca, CMDAT_RES_TYPE1);
if (ret < 0)
goto card_init_error;
ret = sd_command(SD_SEND_CSD, currcard->rca, temp_reg, CMDAT_RES_TYPE2);
if (ret < 0)
goto card_init_error;
for(i=0; i<4; i++)
currcard->csd[i] = temp_reg[3-i];
sd_parse_csd(currcard);
MMC_CLKRT = 0; /* switch to highest clock rate */
ret = sd_command(SD_SELECT_CARD, currcard->rca, NULL,
0x80 | CMDAT_RES_TYPE1);
if (ret < 0)
goto card_init_error;
ret = sd_command(SD_APP_CMD, currcard->rca, NULL, CMDAT_RES_TYPE1);
if (ret < 0)
goto card_init_error;
ret = sd_command(SD_SET_BUS_WIDTH, currcard->rca | 2, NULL,
CMDAT_RES_TYPE1); /* 4 bit */
if (ret < 0)
goto card_init_error;
ret = sd_command(SD_SET_BLOCKLEN, currcard->blocksize, NULL,
CMDAT_RES_TYPE1);
if (ret < 0)
goto card_init_error;
MMC_BLKLEN = currcard->blocksize;
/* If this card is >4GB & not SDHC, then we need to enable bank switching */
if( (currcard->numblocks >= BLOCKS_PER_BANK) &&
((currcard->ocr & (1<<30)) == 0) )
{
MMC_SD_STATE = SD_TRAN;
MMC_NUMBLK = 1;
ret = sd_command(SD_SWITCH_FUNC, 0x80ffffef, NULL,
0x1c00 | CMDAT_DATA_EN | CMDAT_RES_TYPE1);
if (ret < 0)
goto card_init_error;
/* Read 512 bytes from the card.
The first 512 bits contain the status information
TODO: Do something useful with this! */
dataptr = carddata;
for (i = 0; i < SD_BLOCK_SIZE/2; i += FIFO_LEN)
{
/* Wait for the FIFO to be full */
if (sd_poll_status(STAT_RECV_FIFO_FULL, 100000))
{
copy_read_sectors_slow(&dataptr);
continue;
}
ret = -EC_FIFO_ENA_BANK_EMPTY;
goto card_init_error;
}
}
currcard->initialized = 1;
return;
/* Card failed to initialize so disable it */
card_init_error:
currcard->initialized = ret;
}
/* lock must already be aquired */
static void sd_select_device(int card_no)
{
currcard = &card_info[card_no];
if (card_no == 0)
{
/* Main card always gets a chance */
sd_status[0].retry = 0;
}
if (currcard->initialized > 0)
{
/* This card is already initialized - switch to it */
sd_card_mux(card_no);
return;
}
if (currcard->initialized == 0)
{
/* Card needs (re)init */
sd_init_device(card_no);
}
}
/* API Functions */
int sd_read_sectors(IF_MD(int drive,) unsigned long start, int incount,
void* inbuf)
{
#ifndef HAVE_MULTIDRIVE
const int drive = 0;
#endif
int ret;
unsigned char *buf, *buf_end;
unsigned int bank;
/* TODO: Add DMA support. */
mutex_lock(&sd_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
enable_controller(true);
led(true);
sd_read_retry:
if (drive != 0 && !card_detect_target())
{
/* no external sd-card inserted */
ret = -EC_NOCARD;
goto sd_read_error;
}
sd_select_device(drive);
if (currcard->initialized < 0)
{
ret = currcard->initialized;
goto sd_read_error;
}
last_disk_activity = current_tick;
/* Only switch banks with non-SDHC cards */
if((currcard->ocr & (1<<30))==0)
{
bank = start / BLOCKS_PER_BANK;
if (currcard->current_bank != bank)
{
ret = sd_select_bank(bank);
if (ret < 0)
goto sd_read_error;
}
start -= bank * BLOCKS_PER_BANK;
}
ret = sd_wait_for_state(SD_TRAN, EC_TRAN_READ_ENTRY);
if (ret < 0)
goto sd_read_error;
MMC_NUMBLK = incount;
#ifdef HAVE_HOTSWAP
if(currcard->ocr & (1<<30) )
{
/* SDHC */
ret = sd_command(SD_READ_MULTIPLE_BLOCK, start, NULL,
0x1c00 | CMDAT_BUSY | CMDAT_DATA_EN | CMDAT_RES_TYPE1);
}
else
#endif
{
ret = sd_command(SD_READ_MULTIPLE_BLOCK, start * SD_BLOCK_SIZE, NULL,
0x1c00 | CMDAT_BUSY | CMDAT_DATA_EN | CMDAT_RES_TYPE1);
}
if (ret < 0)
goto sd_read_error;
/* TODO: Don't assume SD_BLOCK_SIZE == SECTOR_SIZE */
buf_end = (unsigned char *)inbuf + incount * currcard->blocksize;
for (buf = inbuf; buf < buf_end;)
{
/* Wait for the FIFO to be full */
if (sd_poll_status(STAT_RECV_FIFO_FULL, 0x80000))
{
copy_read_sectors_fast(&buf); /* Copy one chunk of 16 words */
/* TODO: Switch bank if necessary */
continue;
}
ret = -EC_FIFO_READ_FULL;
goto sd_read_error;
}
last_disk_activity = current_tick;
ret = sd_command(SD_STOP_TRANSMISSION, 0, NULL, CMDAT_RES_TYPE1);
if (ret < 0)
goto sd_read_error;
ret = sd_wait_for_state(SD_TRAN, EC_TRAN_READ_EXIT);
if (ret < 0)
goto sd_read_error;
while (1)
{
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
enable_controller(false);
mutex_unlock(&sd_mtx);
return ret;
sd_read_error:
if (sd_status[drive].retry < sd_status[drive].retry_max
&& ret != -EC_NOCARD)
{
sd_status[drive].retry++;
currcard->initialized = 0;
goto sd_read_retry;
}
}
}
int sd_write_sectors(IF_MD(int drive,) unsigned long start, int count,
const void* outbuf)
{
/* Write support is not finished yet */
/* TODO: The standard suggests using ACMD23 prior to writing multiple blocks
to improve performance */
#ifndef HAVE_MULTIDRIVE
const int drive = 0;
#endif
int ret;
const unsigned char *buf, *buf_end;
unsigned int bank;
mutex_lock(&sd_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
enable_controller(true);
led(true);
sd_write_retry:
if (drive != 0 && !card_detect_target())
{
/* no external sd-card inserted */
ret = -EC_NOCARD;
goto sd_write_error;
}
sd_select_device(drive);
if (currcard->initialized < 0)
{
ret = currcard->initialized;
goto sd_write_error;
}
/* Only switch banks with non-SDHC cards */
if((currcard->ocr & (1<<30))==0)
{
bank = start / BLOCKS_PER_BANK;
if (currcard->current_bank != bank)
{
ret = sd_select_bank(bank);
if (ret < 0)
goto sd_write_error;
}
start -= bank * BLOCKS_PER_BANK;
}
check_time[EC_WRITE_TIMEOUT] = USEC_TIMER;
ret = sd_wait_for_state(SD_TRAN, EC_TRAN_WRITE_ENTRY);
if (ret < 0)
goto sd_write_error;
MMC_NUMBLK = count;
#ifdef HAVE_HOTSWAP
if(currcard->ocr & (1<<30) )
{
/* SDHC */
ret = sd_command(SD_WRITE_MULTIPLE_BLOCK, start, NULL,
CMDAT_WR_RD | CMDAT_DATA_EN | CMDAT_RES_TYPE1);
}
else
#endif
{
ret = sd_command(SD_WRITE_MULTIPLE_BLOCK, start*SD_BLOCK_SIZE, NULL,
CMDAT_WR_RD | CMDAT_DATA_EN | CMDAT_RES_TYPE1);
}
if (ret < 0)
goto sd_write_error;
buf_end = outbuf + count * currcard->blocksize - 2*FIFO_LEN;
for (buf = outbuf; buf <= buf_end;)
{
if (buf == buf_end)
{
/* Set MMC_SD_STATE to SD_PRG for the last buffer fill */
MMC_SD_STATE = SD_PRG;
}
copy_write_sectors(&buf); /* Copy one chunk of 16 words */
/* TODO: Switch bank if necessary */
/* Wait for the FIFO to empty */
if (!sd_poll_status(STAT_XMIT_FIFO_EMPTY, 0x80000))
{
ret = -EC_FIFO_WR_EMPTY;
goto sd_write_error;
}
}
last_disk_activity = current_tick;
if (!sd_poll_status(STAT_PRG_DONE, 0x80000))
{
ret = -EC_FIFO_WR_DONE;
goto sd_write_error;
}
ret = sd_command(SD_STOP_TRANSMISSION, 0, NULL, CMDAT_RES_TYPE1);
if (ret < 0)
goto sd_write_error;
ret = sd_wait_for_state(SD_TRAN, EC_TRAN_WRITE_EXIT);
if (ret < 0)
goto sd_write_error;
while (1)
{
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
enable_controller(false);
mutex_unlock(&sd_mtx);
return ret;
sd_write_error:
if (sd_status[drive].retry < sd_status[drive].retry_max
&& ret != -EC_NOCARD)
{
sd_status[drive].retry++;
currcard->initialized = 0;
goto sd_write_retry;
}
}
}
void sd_enable(bool on)
{
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(&sd_mtx);
enable_controller(on);
mutex_unlock(&sd_mtx);
}
int sd_init(void)
{
int 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
mutex_init(&sd_mtx);
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
/* init controller */
#if defined(PHILIPS_SA9200)
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
GPIOA_ENABLE = 0x00;
GPIO_SET_BITWISE(GPIOD_ENABLE, 0x01);
#else
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
outl(inl(0x70000088) & ~(0x4), 0x70000088);
outl(inl(0x7000008c) & ~(0x4), 0x7000008c);
GPO32_ENABLE |= 0x4;
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
GPIO_SET_BITWISE(GPIOG_ENABLE, (0x3 << 5));
GPIO_SET_BITWISE(GPIOG_OUTPUT_EN, (0x3 << 5));
GPIO_SET_BITWISE(GPIOG_OUTPUT_VAL, (0x3 << 5));
#endif
#ifdef HAVE_HOTSWAP
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
/* enable card detection port - mask interrupt first */
#ifdef SANSA_E200
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
GPIO_CLEAR_BITWISE(GPIOA_INT_EN, 0x80);
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
GPIO_CLEAR_BITWISE(GPIOA_OUTPUT_EN, 0x80);
GPIO_SET_BITWISE(GPIOA_ENABLE, 0x80);
#elif defined SANSA_C200
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
GPIO_CLEAR_BITWISE(GPIOL_INT_EN, 0x08);
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
GPIO_CLEAR_BITWISE(GPIOL_OUTPUT_EN, 0x08);
GPIO_SET_BITWISE(GPIOL_ENABLE, 0x08);
#endif
#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
sd_select_device(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
if (currcard->initialized < 0)
ret = currcard->initialized;
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
/* enable interupt for the mSD card */
sleep(HZ/10);
#ifdef HAVE_HOTSWAP
#ifdef SANSA_E200
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
CPU_INT_EN = HI_MASK;
CPU_HI_INT_EN = GPIO0_MASK;
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
GPIOA_INT_LEV = (0x80 << 8) | (~GPIOA_INPUT_VAL & 0x80);
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
GPIOA_INT_CLR = 0x80;
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
/* enable the card detect interrupt */
GPIO_SET_BITWISE(GPIOA_INT_EN, 0x80);
#elif defined SANSA_C200
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
CPU_INT_EN = HI_MASK;
CPU_HI_INT_EN = GPIO2_MASK;
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
GPIOL_INT_LEV = (0x08 << 8) | (~GPIOL_INPUT_VAL & 0x08);
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
GPIOL_INT_CLR = 0x08;
/* enable the card detect interrupt */
GPIO_SET_BITWISE(GPIOL_INT_EN, 0x08);
#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
#endif /* HAVE_HOTSWAP */
mutex_unlock(&sd_mtx);
return ret;
}
tCardInfo *card_get_info_target(int card_no)
{
return &card_info[card_no];
}
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_HOTSWAP
static int sd1_oneshot_callback(struct timeout *tmo)
{
/* This is called only if the state was stable for 300ms - check state
* and post appropriate event. */
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
queue_broadcast(card_detect_target() ? SYS_HOTSWAP_INSERTED :
SYS_HOTSWAP_EXTRACTED,
sd_first_drive+1);
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
(void)tmo;
}
/* called on insertion/removal interrupt */
void microsd_int(void)
{
static struct timeout sd1_oneshot;
#ifdef SANSA_E200
GPIO_CLEAR_BITWISE(GPIOA_INT_EN, 0x80);
GPIOA_INT_LEV = (0x80 << 8) | (~GPIOA_INPUT_VAL & 0x80);
GPIOA_INT_CLR = 0x80;
GPIO_SET_BITWISE(GPIOA_INT_EN, 0x80);
#elif defined SANSA_C200
GPIO_CLEAR_BITWISE(GPIOL_INT_EN, 0x08);
GPIOL_INT_LEV = (0x08 << 8) | (~GPIOL_INPUT_VAL & 0x08);
GPIOL_INT_CLR = 0x08;
GPIO_SET_BITWISE(GPIOL_INT_EN, 0x08);
#endif
timeout_register(&sd1_oneshot, sd1_oneshot_callback, (3*HZ/10), 0);
}
#endif /* HAVE_HOTSWAP */
long sd_last_disk_activity(void)
{
return last_disk_activity;
}
#ifdef HAVE_HOTSWAP
bool sd_removable(IF_MD_NONVOID(int drive))
{
#ifndef HAVE_MULTIDRIVE
const int drive=0;
#endif
return (drive==1);
}
bool sd_present(IF_MD_NONVOID(int drive))
{
#ifndef HAVE_MULTIDRIVE
const int drive=0;
#endif
if(drive==0)
{
return true;
}
else
{
return card_detect_target();
}
}
#endif
#ifdef CONFIG_STORAGE_MULTI
int sd_num_drives(int first_drive)
{
#ifdef HAVE_HOTSWAP
/* Store which logical drive number(s) we have been assigned */
sd_first_drive = first_drive;
#else
(void)first_drive;
#endif
#ifdef HAVE_MULTIDRIVE
return 2;
#else
return 1;
#endif
}
#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 sd_event(long id, intptr_t data)
{
int rc = 0;
switch (id)
{
#ifdef HAVE_HOTSWAP
case SYS_HOTSWAP_INSERTED:
case SYS_HOTSWAP_EXTRACTED:
mutex_lock(&sd_mtx); /* lock-out card activity */
/* Force card init for new card, re-init for re-inserted one or
* clear if the last attempt to init failed with an error. */
card_info[data].initialized = 0;
sd_status[data].retry = 0;
/* Access is now safe */
mutex_unlock(&sd_mtx);
break;
#endif /* HAVE_HOTSWAP */
case Q_STORAGE_TICK:
/* never let a timer wrap confuse us */
next_yield = USEC_TIMER;
default:
rc = storage_event_default_handler(id, data, last_disk_activity,
STORAGE_SD);
break;
}
return rc;
}