2008-10-25 06:18:12 +00:00
|
|
|
/*-
|
|
|
|
* Copyright (c) 2007, Juniper Networks, Inc.
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
* Copyright (c) 2012-2013, SRI International
|
2008-10-25 06:18:12 +00:00
|
|
|
* All rights reserved.
|
|
|
|
*
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
* Portions of this software were developed by SRI International and the
|
|
|
|
* University of Cambridge Computer Laboratory under DARPA/AFRL contract
|
|
|
|
* (FA8750-10-C-0237) ("CTSRD"), as part of the DARPA CRASH research
|
|
|
|
* programme.
|
|
|
|
*
|
2008-10-25 06:18:12 +00:00
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
* modification, are permitted provided that the following conditions
|
|
|
|
* are met:
|
|
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer.
|
|
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
|
|
* documentation and/or other materials provided with the distribution.
|
|
|
|
* 3. Neither the name of the author nor the names of any co-contributors
|
|
|
|
* may be used to endorse or promote products derived from this software
|
|
|
|
* without specific prior written permission.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
|
|
|
|
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
|
|
|
|
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
|
|
|
|
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
|
|
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
|
|
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
|
|
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
|
|
|
|
* AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
|
|
|
|
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
|
|
* SUCH DAMAGE.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
2009-02-05 18:12:07 +00:00
|
|
|
#include "opt_cfi.h"
|
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/systm.h>
|
|
|
|
#include <sys/bus.h>
|
|
|
|
#include <sys/conf.h>
|
2012-03-27 15:13:12 +00:00
|
|
|
#include <sys/endian.h>
|
2013-04-30 18:48:11 +00:00
|
|
|
#include <sys/kenv.h>
|
2008-10-25 06:18:12 +00:00
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
#include <sys/module.h>
|
|
|
|
#include <sys/rman.h>
|
|
|
|
#include <sys/sysctl.h>
|
|
|
|
|
|
|
|
#include <machine/bus.h>
|
|
|
|
|
|
|
|
#include <dev/cfi/cfi_reg.h>
|
|
|
|
#include <dev/cfi/cfi_var.h>
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
static void cfi_add_sysctls(struct cfi_softc *);
|
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
extern struct cdevsw cfi_cdevsw;
|
|
|
|
|
|
|
|
char cfi_driver_name[] = "cfi";
|
|
|
|
devclass_t cfi_devclass;
|
2009-03-09 23:16:02 +00:00
|
|
|
devclass_t cfi_diskclass;
|
2008-10-25 06:18:12 +00:00
|
|
|
|
|
|
|
uint32_t
|
2012-03-27 15:13:12 +00:00
|
|
|
cfi_read_raw(struct cfi_softc *sc, u_int ofs)
|
2008-10-25 06:18:12 +00:00
|
|
|
{
|
|
|
|
uint32_t val;
|
|
|
|
|
|
|
|
ofs &= ~(sc->sc_width - 1);
|
|
|
|
switch (sc->sc_width) {
|
|
|
|
case 1:
|
|
|
|
val = bus_space_read_1(sc->sc_tag, sc->sc_handle, ofs);
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
val = bus_space_read_2(sc->sc_tag, sc->sc_handle, ofs);
|
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
val = bus_space_read_4(sc->sc_tag, sc->sc_handle, ofs);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
val = ~0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return (val);
|
|
|
|
}
|
|
|
|
|
2012-03-27 15:13:12 +00:00
|
|
|
uint32_t
|
|
|
|
cfi_read(struct cfi_softc *sc, u_int ofs)
|
|
|
|
{
|
|
|
|
uint32_t val;
|
|
|
|
uint16_t sval;
|
|
|
|
|
|
|
|
ofs &= ~(sc->sc_width - 1);
|
|
|
|
switch (sc->sc_width) {
|
|
|
|
case 1:
|
|
|
|
val = bus_space_read_1(sc->sc_tag, sc->sc_handle, ofs);
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
sval = bus_space_read_2(sc->sc_tag, sc->sc_handle, ofs);
|
|
|
|
val = le16toh(sval);
|
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
val = bus_space_read_4(sc->sc_tag, sc->sc_handle, ofs);
|
|
|
|
val = le32toh(val);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
val = ~0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return (val);
|
|
|
|
}
|
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
static void
|
|
|
|
cfi_write(struct cfi_softc *sc, u_int ofs, u_int val)
|
|
|
|
{
|
|
|
|
|
|
|
|
ofs &= ~(sc->sc_width - 1);
|
|
|
|
switch (sc->sc_width) {
|
|
|
|
case 1:
|
|
|
|
bus_space_write_1(sc->sc_tag, sc->sc_handle, ofs, val);
|
|
|
|
break;
|
|
|
|
case 2:
|
2012-03-27 15:13:12 +00:00
|
|
|
bus_space_write_2(sc->sc_tag, sc->sc_handle, ofs, htole16(val));
|
2008-10-25 06:18:12 +00:00
|
|
|
break;
|
|
|
|
case 4:
|
2012-03-27 15:13:12 +00:00
|
|
|
bus_space_write_4(sc->sc_tag, sc->sc_handle, ofs, htole32(val));
|
2008-10-25 06:18:12 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
uint8_t
|
|
|
|
cfi_read_qry(struct cfi_softc *sc, u_int ofs)
|
|
|
|
{
|
|
|
|
uint8_t val;
|
|
|
|
|
|
|
|
cfi_write(sc, CFI_QRY_CMD_ADDR * sc->sc_width, CFI_QRY_CMD_DATA);
|
|
|
|
val = cfi_read(sc, ofs * sc->sc_width);
|
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
|
|
|
return (val);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
cfi_amd_write(struct cfi_softc *sc, u_int ofs, u_int addr, u_int data)
|
|
|
|
{
|
|
|
|
|
|
|
|
cfi_write(sc, ofs + AMD_ADDR_START, CFI_AMD_UNLOCK);
|
|
|
|
cfi_write(sc, ofs + AMD_ADDR_ACK, CFI_AMD_UNLOCK_ACK);
|
|
|
|
cfi_write(sc, ofs + addr, data);
|
|
|
|
}
|
|
|
|
|
|
|
|
static char *
|
|
|
|
cfi_fmtsize(uint32_t sz)
|
|
|
|
{
|
|
|
|
static char buf[8];
|
|
|
|
static const char *sfx[] = { "", "K", "M", "G" };
|
|
|
|
int sfxidx;
|
|
|
|
|
|
|
|
sfxidx = 0;
|
|
|
|
while (sfxidx < 3 && sz > 1023) {
|
|
|
|
sz /= 1024;
|
|
|
|
sfxidx++;
|
|
|
|
}
|
|
|
|
|
|
|
|
sprintf(buf, "%u%sB", sz, sfx[sfxidx]);
|
|
|
|
return (buf);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
cfi_probe(device_t dev)
|
|
|
|
{
|
|
|
|
char desc[80];
|
|
|
|
struct cfi_softc *sc;
|
|
|
|
char *vend_str;
|
|
|
|
int error;
|
|
|
|
uint16_t iface, vend;
|
|
|
|
|
|
|
|
sc = device_get_softc(dev);
|
|
|
|
sc->sc_dev = dev;
|
|
|
|
|
|
|
|
sc->sc_rid = 0;
|
|
|
|
sc->sc_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->sc_rid,
|
|
|
|
RF_ACTIVE);
|
|
|
|
if (sc->sc_res == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
|
|
|
|
sc->sc_tag = rman_get_bustag(sc->sc_res);
|
|
|
|
sc->sc_handle = rman_get_bushandle(sc->sc_res);
|
|
|
|
|
2009-02-03 19:09:16 +00:00
|
|
|
if (sc->sc_width == 0) {
|
|
|
|
sc->sc_width = 1;
|
|
|
|
while (sc->sc_width <= 4) {
|
|
|
|
if (cfi_read_qry(sc, CFI_QRY_IDENT) == 'Q')
|
|
|
|
break;
|
|
|
|
sc->sc_width <<= 1;
|
|
|
|
}
|
|
|
|
} else if (cfi_read_qry(sc, CFI_QRY_IDENT) != 'Q') {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out;
|
2008-10-25 06:18:12 +00:00
|
|
|
}
|
|
|
|
if (sc->sc_width > 4) {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We got a Q. Check if we also have the R and the Y. */
|
|
|
|
if (cfi_read_qry(sc, CFI_QRY_IDENT + 1) != 'R' ||
|
|
|
|
cfi_read_qry(sc, CFI_QRY_IDENT + 2) != 'Y') {
|
|
|
|
error = ENXIO;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Get the vendor and command set. */
|
|
|
|
vend = cfi_read_qry(sc, CFI_QRY_VEND) |
|
|
|
|
(cfi_read_qry(sc, CFI_QRY_VEND + 1) << 8);
|
|
|
|
|
|
|
|
sc->sc_cmdset = vend;
|
|
|
|
|
|
|
|
switch (vend) {
|
|
|
|
case CFI_VEND_AMD_ECS:
|
|
|
|
case CFI_VEND_AMD_SCS:
|
|
|
|
vend_str = "AMD/Fujitsu";
|
|
|
|
break;
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
vend_str = "Intel/Sharp";
|
|
|
|
break;
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
|
|
|
vend_str = "Intel";
|
|
|
|
break;
|
|
|
|
case CFI_VEND_MITSUBISHI_ECS:
|
|
|
|
case CFI_VEND_MITSUBISHI_SCS:
|
|
|
|
vend_str = "Mitsubishi";
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
vend_str = "Unknown vendor";
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Get the device size. */
|
|
|
|
sc->sc_size = 1U << cfi_read_qry(sc, CFI_QRY_SIZE);
|
|
|
|
|
|
|
|
/* Sanity-check the I/F */
|
|
|
|
iface = cfi_read_qry(sc, CFI_QRY_IFACE) |
|
|
|
|
(cfi_read_qry(sc, CFI_QRY_IFACE + 1) << 8);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Adding 1 to iface will give us a bit-wise "switch"
|
|
|
|
* that allows us to test for the interface width by
|
|
|
|
* testing a single bit.
|
|
|
|
*/
|
|
|
|
iface++;
|
|
|
|
|
|
|
|
error = (iface & sc->sc_width) ? 0 : EINVAL;
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
snprintf(desc, sizeof(desc), "%s - %s", vend_str,
|
|
|
|
cfi_fmtsize(sc->sc_size));
|
|
|
|
device_set_desc_copy(dev, desc);
|
|
|
|
|
|
|
|
out:
|
|
|
|
bus_release_resource(dev, SYS_RES_MEMORY, sc->sc_rid, sc->sc_res);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
cfi_attach(device_t dev)
|
|
|
|
{
|
|
|
|
struct cfi_softc *sc;
|
|
|
|
u_int blksz, blocks;
|
|
|
|
u_int r, u;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
uint64_t mtoexp, ttoexp;
|
2013-04-30 18:48:11 +00:00
|
|
|
#ifdef CFI_SUPPORT_STRATAFLASH
|
|
|
|
uint64_t ppr;
|
|
|
|
char name[KENV_MNAMELEN], value[32];
|
|
|
|
#endif
|
2008-10-25 06:18:12 +00:00
|
|
|
|
|
|
|
sc = device_get_softc(dev);
|
|
|
|
sc->sc_dev = dev;
|
|
|
|
|
|
|
|
sc->sc_rid = 0;
|
|
|
|
sc->sc_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->sc_rid,
|
2013-10-18 20:52:42 +00:00
|
|
|
#ifndef ATSE_CFI_HACK
|
2008-10-25 06:18:12 +00:00
|
|
|
RF_ACTIVE);
|
2013-10-18 20:52:42 +00:00
|
|
|
#else
|
|
|
|
RF_ACTIVE | RF_SHAREABLE);
|
|
|
|
#endif
|
2008-10-25 06:18:12 +00:00
|
|
|
if (sc->sc_res == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
|
|
|
|
sc->sc_tag = rman_get_bustag(sc->sc_res);
|
|
|
|
sc->sc_handle = rman_get_bushandle(sc->sc_res);
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
/* Get time-out values for erase, write, and buffer write. */
|
|
|
|
ttoexp = cfi_read_qry(sc, CFI_QRY_TTO_ERASE);
|
|
|
|
mtoexp = cfi_read_qry(sc, CFI_QRY_MTO_ERASE);
|
|
|
|
if (ttoexp == 0) {
|
|
|
|
device_printf(dev, "erase timeout == 0, using 2^16ms\n");
|
|
|
|
ttoexp = 16;
|
|
|
|
}
|
|
|
|
if (ttoexp > 41) {
|
|
|
|
device_printf(dev, "insane timeout: 2^%jdms\n", ttoexp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
if (mtoexp == 0) {
|
|
|
|
device_printf(dev, "max erase timeout == 0, using 2^%jdms\n",
|
|
|
|
ttoexp + 4);
|
|
|
|
mtoexp = 4;
|
|
|
|
}
|
|
|
|
if (ttoexp + mtoexp > 41) {
|
|
|
|
device_printf(dev, "insane max erase timeout: 2^%jd\n",
|
|
|
|
ttoexp + mtoexp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
sc->sc_typical_timeouts[CFI_TIMEOUT_ERASE] = SBT_1MS * (1ULL << ttoexp);
|
|
|
|
sc->sc_max_timeouts[CFI_TIMEOUT_ERASE] =
|
|
|
|
sc->sc_typical_timeouts[CFI_TIMEOUT_ERASE] * (1ULL << mtoexp);
|
|
|
|
|
|
|
|
ttoexp = cfi_read_qry(sc, CFI_QRY_TTO_WRITE);
|
|
|
|
mtoexp = cfi_read_qry(sc, CFI_QRY_MTO_WRITE);
|
|
|
|
if (ttoexp == 0) {
|
|
|
|
device_printf(dev, "write timeout == 0, using 2^18ns\n");
|
|
|
|
ttoexp = 18;
|
|
|
|
}
|
|
|
|
if (ttoexp > 51) {
|
|
|
|
device_printf(dev, "insane write timeout: 2^%jdus\n", ttoexp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
if (mtoexp == 0) {
|
|
|
|
device_printf(dev, "max write timeout == 0, using 2^%jdms\n",
|
|
|
|
ttoexp + 4);
|
|
|
|
mtoexp = 4;
|
|
|
|
}
|
|
|
|
if (ttoexp + mtoexp > 51) {
|
|
|
|
device_printf(dev, "insane max write timeout: 2^%jdus\n",
|
|
|
|
ttoexp + mtoexp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
sc->sc_typical_timeouts[CFI_TIMEOUT_WRITE] = SBT_1US * (1ULL << ttoexp);
|
|
|
|
sc->sc_max_timeouts[CFI_TIMEOUT_WRITE] =
|
|
|
|
sc->sc_typical_timeouts[CFI_TIMEOUT_WRITE] * (1ULL << mtoexp);
|
|
|
|
|
|
|
|
ttoexp = cfi_read_qry(sc, CFI_QRY_TTO_BUFWRITE);
|
|
|
|
mtoexp = cfi_read_qry(sc, CFI_QRY_MTO_BUFWRITE);
|
|
|
|
/* Don't check for 0, it means not-supported. */
|
|
|
|
if (ttoexp > 51) {
|
|
|
|
device_printf(dev, "insane write timeout: 2^%jdus\n", ttoexp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
if (ttoexp + mtoexp > 51) {
|
|
|
|
device_printf(dev, "insane max write timeout: 2^%jdus\n",
|
|
|
|
ttoexp + mtoexp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
sc->sc_typical_timeouts[CFI_TIMEOUT_BUFWRITE] =
|
|
|
|
SBT_1US * (1ULL << cfi_read_qry(sc, CFI_QRY_TTO_BUFWRITE));
|
|
|
|
sc->sc_max_timeouts[CFI_TIMEOUT_BUFWRITE] =
|
|
|
|
sc->sc_typical_timeouts[CFI_TIMEOUT_BUFWRITE] *
|
|
|
|
(1ULL << cfi_read_qry(sc, CFI_QRY_MTO_BUFWRITE));
|
|
|
|
|
|
|
|
/* Get the maximum size of a multibyte program */
|
|
|
|
if (sc->sc_typical_timeouts[CFI_TIMEOUT_BUFWRITE] != 0)
|
|
|
|
sc->sc_maxbuf = 1 << (cfi_read_qry(sc, CFI_QRY_MAXBUF) |
|
|
|
|
cfi_read_qry(sc, CFI_QRY_MAXBUF) << 8);
|
|
|
|
else
|
|
|
|
sc->sc_maxbuf = 0;
|
2008-10-25 06:18:12 +00:00
|
|
|
|
|
|
|
/* Get erase regions. */
|
|
|
|
sc->sc_regions = cfi_read_qry(sc, CFI_QRY_NREGIONS);
|
|
|
|
sc->sc_region = malloc(sc->sc_regions * sizeof(struct cfi_region),
|
|
|
|
M_TEMP, M_WAITOK | M_ZERO);
|
|
|
|
for (r = 0; r < sc->sc_regions; r++) {
|
|
|
|
blocks = cfi_read_qry(sc, CFI_QRY_REGION(r)) |
|
|
|
|
(cfi_read_qry(sc, CFI_QRY_REGION(r) + 1) << 8);
|
|
|
|
sc->sc_region[r].r_blocks = blocks + 1;
|
|
|
|
|
|
|
|
blksz = cfi_read_qry(sc, CFI_QRY_REGION(r) + 2) |
|
|
|
|
(cfi_read_qry(sc, CFI_QRY_REGION(r) + 3) << 8);
|
|
|
|
sc->sc_region[r].r_blksz = (blksz == 0) ? 128 :
|
|
|
|
blksz * 256;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Reset the device to a default state. */
|
|
|
|
cfi_write(sc, 0, CFI_BCS_CLEAR_STATUS);
|
|
|
|
|
|
|
|
if (bootverbose) {
|
|
|
|
device_printf(dev, "[");
|
|
|
|
for (r = 0; r < sc->sc_regions; r++) {
|
|
|
|
printf("%ux%s%s", sc->sc_region[r].r_blocks,
|
|
|
|
cfi_fmtsize(sc->sc_region[r].r_blksz),
|
|
|
|
(r == sc->sc_regions - 1) ? "]\n" : ",");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
u = device_get_unit(dev);
|
|
|
|
sc->sc_nod = make_dev(&cfi_cdevsw, u, UID_ROOT, GID_WHEEL, 0600,
|
|
|
|
"%s%u", cfi_driver_name, u);
|
|
|
|
sc->sc_nod->si_drv1 = sc;
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
cfi_add_sysctls(sc);
|
|
|
|
|
2013-04-30 18:48:11 +00:00
|
|
|
#ifdef CFI_SUPPORT_STRATAFLASH
|
|
|
|
/*
|
|
|
|
* Store the Intel factory PPR in the environment. In some
|
|
|
|
* cases it is the most unique ID on a board.
|
|
|
|
*/
|
|
|
|
if (cfi_intel_get_factory_pr(sc, &ppr) == 0) {
|
|
|
|
if (snprintf(name, sizeof(name), "%s.factory_ppr",
|
|
|
|
device_get_nameunit(dev)) < (sizeof(name) - 1) &&
|
|
|
|
snprintf(value, sizeof(value), "0x%016jx", ppr) <
|
|
|
|
(sizeof(value) - 1))
|
|
|
|
(void) setenv(name, value);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2009-06-10 17:41:24 +00:00
|
|
|
device_add_child(dev, "cfid", -1);
|
2009-03-09 23:16:02 +00:00
|
|
|
bus_generic_attach(dev);
|
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
static void
|
|
|
|
cfi_add_sysctls(struct cfi_softc *sc)
|
|
|
|
{
|
|
|
|
struct sysctl_ctx_list *ctx;
|
|
|
|
struct sysctl_oid_list *children;
|
|
|
|
|
|
|
|
ctx = device_get_sysctl_ctx(sc->sc_dev);
|
|
|
|
children = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->sc_dev));
|
|
|
|
|
|
|
|
SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
|
|
|
|
"typical_erase_timout_count",
|
|
|
|
CTLFLAG_RD, &sc->sc_tto_counts[CFI_TIMEOUT_ERASE],
|
|
|
|
0, "Number of times the typical erase timeout was exceeded");
|
|
|
|
SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
|
|
|
|
"max_erase_timout_count",
|
|
|
|
CTLFLAG_RD, &sc->sc_mto_counts[CFI_TIMEOUT_ERASE], 0,
|
|
|
|
"Number of times the maximum erase timeout was exceeded");
|
|
|
|
SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
|
|
|
|
"typical_write_timout_count",
|
|
|
|
CTLFLAG_RD, &sc->sc_tto_counts[CFI_TIMEOUT_WRITE], 0,
|
|
|
|
"Number of times the typical write timeout was exceeded");
|
|
|
|
SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
|
|
|
|
"max_write_timout_count",
|
|
|
|
CTLFLAG_RD, &sc->sc_mto_counts[CFI_TIMEOUT_WRITE], 0,
|
|
|
|
"Number of times the maximum write timeout was exceeded");
|
|
|
|
if (sc->sc_maxbuf > 0) {
|
|
|
|
SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
|
|
|
|
"typical_bufwrite_timout_count",
|
|
|
|
CTLFLAG_RD, &sc->sc_tto_counts[CFI_TIMEOUT_BUFWRITE], 0,
|
|
|
|
"Number of times the typical buffered write timeout was "
|
|
|
|
"exceeded");
|
|
|
|
SYSCTL_ADD_UINT(ctx, children, OID_AUTO,
|
|
|
|
"max_bufwrite_timout_count",
|
|
|
|
CTLFLAG_RD, &sc->sc_mto_counts[CFI_TIMEOUT_BUFWRITE], 0,
|
|
|
|
"Number of times the maximum buffered write timeout was "
|
|
|
|
"exceeded");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
int
|
|
|
|
cfi_detach(device_t dev)
|
|
|
|
{
|
|
|
|
struct cfi_softc *sc;
|
|
|
|
|
|
|
|
sc = device_get_softc(dev);
|
|
|
|
|
|
|
|
destroy_dev(sc->sc_nod);
|
|
|
|
free(sc->sc_region, M_TEMP);
|
|
|
|
bus_release_resource(dev, SYS_RES_MEMORY, sc->sc_rid, sc->sc_res);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
cfi_wait_ready(struct cfi_softc *sc, u_int ofs, sbintime_t start,
|
|
|
|
enum cfi_wait_cmd cmd)
|
2008-10-25 06:18:12 +00:00
|
|
|
{
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
int done, error, tto_exceeded;
|
2009-02-05 18:12:07 +00:00
|
|
|
uint32_t st0 = 0, st = 0;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
sbintime_t now;
|
2008-10-25 06:18:12 +00:00
|
|
|
|
|
|
|
done = 0;
|
|
|
|
error = 0;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
tto_exceeded = 0;
|
|
|
|
while (!done && !error) {
|
|
|
|
/*
|
|
|
|
* Save time before we start so we always do one check
|
|
|
|
* after the timeout has expired.
|
|
|
|
*/
|
|
|
|
now = sbinuptime();
|
2008-10-25 06:18:12 +00:00
|
|
|
|
|
|
|
switch (sc->sc_cmdset) {
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
2009-02-05 18:12:07 +00:00
|
|
|
st = cfi_read(sc, ofs);
|
|
|
|
done = (st & CFI_INTEL_STATUS_WSMS);
|
2008-10-25 06:18:12 +00:00
|
|
|
if (done) {
|
2009-02-05 18:12:07 +00:00
|
|
|
/* NB: bit 0 is reserved */
|
|
|
|
st &= ~(CFI_INTEL_XSTATUS_RSVD |
|
|
|
|
CFI_INTEL_STATUS_WSMS |
|
|
|
|
CFI_INTEL_STATUS_RSVD);
|
|
|
|
if (st & CFI_INTEL_STATUS_DPS)
|
2008-10-25 06:18:12 +00:00
|
|
|
error = EPERM;
|
2009-02-05 18:12:07 +00:00
|
|
|
else if (st & CFI_INTEL_STATUS_PSLBS)
|
2008-10-25 06:18:12 +00:00
|
|
|
error = EIO;
|
2009-02-05 18:12:07 +00:00
|
|
|
else if (st & CFI_INTEL_STATUS_ECLBS)
|
2008-10-25 06:18:12 +00:00
|
|
|
error = ENXIO;
|
2009-02-05 18:12:07 +00:00
|
|
|
else if (st)
|
|
|
|
error = EACCES;
|
2008-10-25 06:18:12 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
case CFI_VEND_AMD_SCS:
|
|
|
|
case CFI_VEND_AMD_ECS:
|
2009-02-05 18:12:07 +00:00
|
|
|
st0 = cfi_read(sc, ofs);
|
|
|
|
st = cfi_read(sc, ofs);
|
2008-10-25 06:18:12 +00:00
|
|
|
done = ((st & 0x40) == (st0 & 0x40)) ? 1 : 0;
|
|
|
|
break;
|
|
|
|
}
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
|
|
|
|
if (tto_exceeded ||
|
|
|
|
now > start + sc->sc_typical_timeouts[cmd]) {
|
|
|
|
if (!tto_exceeded) {
|
|
|
|
tto_exceeded = 1;
|
|
|
|
sc->sc_tto_counts[cmd]++;
|
|
|
|
#ifdef CFI_DEBUG_TIMEOUT
|
|
|
|
device_printf(sc->sc_dev,
|
|
|
|
"typical timeout exceeded (cmd %d)", cmd);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
if (now > start + sc->sc_max_timeouts[cmd]) {
|
|
|
|
sc->sc_mto_counts[cmd]++;
|
|
|
|
#ifdef CFI_DEBUG_TIMEOUT
|
|
|
|
device_printf(sc->sc_dev,
|
|
|
|
"max timeout exceeded (cmd %d)", cmd);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
}
|
2008-10-25 06:18:12 +00:00
|
|
|
}
|
|
|
|
if (!done && !error)
|
|
|
|
error = ETIMEDOUT;
|
|
|
|
if (error)
|
2009-02-05 18:12:07 +00:00
|
|
|
printf("\nerror=%d (st 0x%x st0 0x%x)\n", error, st, st0);
|
2008-10-25 06:18:12 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
cfi_write_block(struct cfi_softc *sc)
|
|
|
|
{
|
|
|
|
union {
|
|
|
|
uint8_t *x8;
|
|
|
|
uint16_t *x16;
|
|
|
|
uint32_t *x32;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
} ptr, cpyprt;
|
2008-10-25 06:18:12 +00:00
|
|
|
register_t intr;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
int error, i, neederase = 0;
|
|
|
|
uint32_t st;
|
|
|
|
u_int wlen;
|
|
|
|
sbintime_t start;
|
2008-10-25 06:18:12 +00:00
|
|
|
|
2013-05-30 01:22:50 +00:00
|
|
|
/* Intel flash must be unlocked before modification */
|
|
|
|
switch (sc->sc_cmdset) {
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_INTEL_LBS);
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_INTEL_UB);
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_BCS_READ_ARRAY);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
/* Check if an erase is required. */
|
|
|
|
for (i = 0; i < sc->sc_wrbufsz; i++)
|
|
|
|
if ((sc->sc_wrbuf[i] & sc->sc_wrbufcpy[i]) != sc->sc_wrbuf[i]) {
|
|
|
|
neederase = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (neederase) {
|
|
|
|
intr = intr_disable();
|
|
|
|
start = sbinuptime();
|
|
|
|
/* Erase the block. */
|
|
|
|
switch (sc->sc_cmdset) {
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_BCS_BLOCK_ERASE);
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_BCS_CONFIRM);
|
|
|
|
break;
|
|
|
|
case CFI_VEND_AMD_SCS:
|
|
|
|
case CFI_VEND_AMD_ECS:
|
|
|
|
cfi_amd_write(sc, sc->sc_wrofs, AMD_ADDR_START,
|
|
|
|
CFI_AMD_ERASE_SECTOR);
|
|
|
|
cfi_amd_write(sc, sc->sc_wrofs, 0, CFI_AMD_BLOCK_ERASE);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
/* Better safe than sorry... */
|
|
|
|
intr_restore(intr);
|
|
|
|
return (ENODEV);
|
|
|
|
}
|
|
|
|
intr_restore(intr);
|
|
|
|
error = cfi_wait_ready(sc, sc->sc_wrofs, start,
|
|
|
|
CFI_TIMEOUT_ERASE);
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
} else
|
|
|
|
error = 0;
|
2008-10-25 06:18:12 +00:00
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
/* Write the block using a multibyte write if supported. */
|
2008-10-25 06:18:12 +00:00
|
|
|
ptr.x8 = sc->sc_wrbuf;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
cpyprt.x8 = sc->sc_wrbufcpy;
|
|
|
|
if (sc->sc_maxbuf > sc->sc_width) {
|
|
|
|
switch (sc->sc_cmdset) {
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
|
|
|
for (i = 0; i < sc->sc_wrbufsz; i += wlen) {
|
|
|
|
wlen = MIN(sc->sc_maxbuf, sc->sc_wrbufsz - i);
|
|
|
|
|
|
|
|
intr = intr_disable();
|
|
|
|
|
|
|
|
start = sbinuptime();
|
|
|
|
do {
|
|
|
|
cfi_write(sc, sc->sc_wrofs + i,
|
|
|
|
CFI_BCS_BUF_PROG_SETUP);
|
|
|
|
if (sbinuptime() > start + sc->sc_max_timeouts[CFI_TIMEOUT_BUFWRITE]) {
|
|
|
|
error = ETIMEDOUT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
st = cfi_read(sc, sc->sc_wrofs + i);
|
|
|
|
} while (! (st & CFI_INTEL_STATUS_WSMS));
|
|
|
|
|
|
|
|
cfi_write(sc, sc->sc_wrofs + i,
|
|
|
|
(wlen / sc->sc_width) - 1);
|
|
|
|
switch (sc->sc_width) {
|
|
|
|
case 1:
|
|
|
|
bus_space_write_region_1(sc->sc_tag,
|
|
|
|
sc->sc_handle, sc->sc_wrofs + i,
|
|
|
|
ptr.x8 + i, wlen);
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
bus_space_write_region_2(sc->sc_tag,
|
|
|
|
sc->sc_handle, sc->sc_wrofs + i,
|
|
|
|
ptr.x16 + i / 2, wlen / 2);
|
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
bus_space_write_region_4(sc->sc_tag,
|
|
|
|
sc->sc_handle, sc->sc_wrofs + i,
|
|
|
|
ptr.x32 + i / 4, wlen / 4);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
cfi_write(sc, sc->sc_wrofs + i,
|
|
|
|
CFI_BCS_CONFIRM);
|
|
|
|
|
|
|
|
intr_restore(intr);
|
|
|
|
|
|
|
|
error = cfi_wait_ready(sc, sc->sc_wrofs + i,
|
|
|
|
start, CFI_TIMEOUT_BUFWRITE);
|
|
|
|
if (error != 0)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
goto out;
|
|
|
|
default:
|
|
|
|
/* Fall through to single word case */
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Write the block one byte/word at a time. */
|
2008-10-25 06:18:12 +00:00
|
|
|
for (i = 0; i < sc->sc_wrbufsz; i += sc->sc_width) {
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
/* Avoid writing unless we are actually changing bits */
|
|
|
|
if (!neederase) {
|
|
|
|
switch (sc->sc_width) {
|
|
|
|
case 1:
|
|
|
|
if(*(ptr.x8 + i) == *(cpyprt.x8 + i))
|
|
|
|
continue;
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
if(*(ptr.x16 + i / 2) == *(cpyprt.x16 + i / 2))
|
|
|
|
continue;
|
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
if(*(ptr.x32 + i / 4) == *(cpyprt.x32 + i / 4))
|
|
|
|
continue;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
/*
|
|
|
|
* Make sure the command to start a write and the
|
|
|
|
* actual write happens back-to-back without any
|
|
|
|
* excessive delays.
|
|
|
|
*/
|
|
|
|
intr = intr_disable();
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
start = sbinuptime();
|
2008-10-25 06:18:12 +00:00
|
|
|
switch (sc->sc_cmdset) {
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
|
|
|
cfi_write(sc, sc->sc_wrofs + i, CFI_BCS_PROGRAM);
|
|
|
|
break;
|
|
|
|
case CFI_VEND_AMD_SCS:
|
|
|
|
case CFI_VEND_AMD_ECS:
|
|
|
|
cfi_amd_write(sc, 0, AMD_ADDR_START, CFI_AMD_PROGRAM);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
switch (sc->sc_width) {
|
|
|
|
case 1:
|
|
|
|
bus_space_write_1(sc->sc_tag, sc->sc_handle,
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
sc->sc_wrofs + i, *(ptr.x8 + i));
|
2008-10-25 06:18:12 +00:00
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
bus_space_write_2(sc->sc_tag, sc->sc_handle,
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
sc->sc_wrofs + i, *(ptr.x16 + i / 2));
|
2008-10-25 06:18:12 +00:00
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
bus_space_write_4(sc->sc_tag, sc->sc_handle,
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
sc->sc_wrofs + i, *(ptr.x32 + i / 4));
|
2008-10-25 06:18:12 +00:00
|
|
|
break;
|
|
|
|
}
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
|
2008-10-25 06:18:12 +00:00
|
|
|
intr_restore(intr);
|
|
|
|
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
error = cfi_wait_ready(sc, sc->sc_wrofs, start,
|
|
|
|
CFI_TIMEOUT_WRITE);
|
2008-10-25 06:18:12 +00:00
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* error is 0. */
|
|
|
|
|
|
|
|
out:
|
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
2013-05-30 01:22:50 +00:00
|
|
|
|
|
|
|
/* Relock Intel flash */
|
|
|
|
switch (sc->sc_cmdset) {
|
|
|
|
case CFI_VEND_INTEL_ECS:
|
|
|
|
case CFI_VEND_INTEL_SCS:
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_INTEL_LBS);
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_INTEL_LB);
|
|
|
|
cfi_write(sc, sc->sc_wrofs, CFI_BCS_READ_ARRAY);
|
|
|
|
break;
|
|
|
|
}
|
2008-10-25 06:18:12 +00:00
|
|
|
return (error);
|
|
|
|
}
|
2009-02-05 18:12:07 +00:00
|
|
|
|
|
|
|
#ifdef CFI_SUPPORT_STRATAFLASH
|
|
|
|
/*
|
|
|
|
* Intel StrataFlash Protection Register Support.
|
|
|
|
*
|
|
|
|
* The memory includes a 128-bit Protection Register that can be
|
|
|
|
* used for security. There are two 64-bit segments; one is programmed
|
|
|
|
* at the factory with a unique 64-bit number which is immutable.
|
|
|
|
* The other segment is left blank for User (OEM) programming.
|
2009-02-07 05:03:25 +00:00
|
|
|
* The User/OEM segment is One Time Programmable (OTP). It can also
|
2009-02-08 20:29:37 +00:00
|
|
|
* be locked to prevent any further writes by setting bit 0 of the
|
2009-02-07 05:03:25 +00:00
|
|
|
* Protection Lock Register (PLR). The PLR can written only once.
|
2009-02-05 18:12:07 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
static uint16_t
|
|
|
|
cfi_get16(struct cfi_softc *sc, int off)
|
|
|
|
{
|
|
|
|
uint16_t v = bus_space_read_2(sc->sc_tag, sc->sc_handle, off<<1);
|
|
|
|
return v;
|
|
|
|
}
|
|
|
|
|
2009-02-07 05:32:19 +00:00
|
|
|
#ifdef CFI_ARMEDANDDANGEROUS
|
2009-02-05 18:12:07 +00:00
|
|
|
static void
|
|
|
|
cfi_put16(struct cfi_softc *sc, int off, uint16_t v)
|
|
|
|
{
|
|
|
|
bus_space_write_2(sc->sc_tag, sc->sc_handle, off<<1, v);
|
|
|
|
}
|
2009-02-07 05:32:19 +00:00
|
|
|
#endif
|
2009-02-05 18:12:07 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Read the factory-defined 64-bit segment of the PR.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
cfi_intel_get_factory_pr(struct cfi_softc *sc, uint64_t *id)
|
|
|
|
{
|
|
|
|
if (sc->sc_cmdset != CFI_VEND_INTEL_ECS)
|
|
|
|
return EOPNOTSUPP;
|
|
|
|
KASSERT(sc->sc_width == 2, ("sc_width %d", sc->sc_width));
|
|
|
|
|
|
|
|
cfi_write(sc, 0, CFI_INTEL_READ_ID);
|
|
|
|
*id = ((uint64_t)cfi_get16(sc, CFI_INTEL_PR(0)))<<48 |
|
|
|
|
((uint64_t)cfi_get16(sc, CFI_INTEL_PR(1)))<<32 |
|
|
|
|
((uint64_t)cfi_get16(sc, CFI_INTEL_PR(2)))<<16 |
|
|
|
|
((uint64_t)cfi_get16(sc, CFI_INTEL_PR(3)));
|
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read the User/OEM 64-bit segment of the PR.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
cfi_intel_get_oem_pr(struct cfi_softc *sc, uint64_t *id)
|
|
|
|
{
|
|
|
|
if (sc->sc_cmdset != CFI_VEND_INTEL_ECS)
|
|
|
|
return EOPNOTSUPP;
|
|
|
|
KASSERT(sc->sc_width == 2, ("sc_width %d", sc->sc_width));
|
|
|
|
|
|
|
|
cfi_write(sc, 0, CFI_INTEL_READ_ID);
|
|
|
|
*id = ((uint64_t)cfi_get16(sc, CFI_INTEL_PR(4)))<<48 |
|
|
|
|
((uint64_t)cfi_get16(sc, CFI_INTEL_PR(5)))<<32 |
|
|
|
|
((uint64_t)cfi_get16(sc, CFI_INTEL_PR(6)))<<16 |
|
|
|
|
((uint64_t)cfi_get16(sc, CFI_INTEL_PR(7)));
|
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Write the User/OEM 64-bit segment of the PR.
|
2009-02-07 05:03:25 +00:00
|
|
|
* XXX should allow writing individual words/bytes
|
2009-02-05 18:12:07 +00:00
|
|
|
*/
|
|
|
|
int
|
|
|
|
cfi_intel_set_oem_pr(struct cfi_softc *sc, uint64_t id)
|
|
|
|
{
|
2009-02-07 05:03:25 +00:00
|
|
|
#ifdef CFI_ARMEDANDDANGEROUS
|
2009-02-05 18:12:07 +00:00
|
|
|
register_t intr;
|
|
|
|
int i, error;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
sbintime_t start;
|
2009-02-07 05:03:25 +00:00
|
|
|
#endif
|
2009-02-05 18:12:07 +00:00
|
|
|
|
|
|
|
if (sc->sc_cmdset != CFI_VEND_INTEL_ECS)
|
|
|
|
return EOPNOTSUPP;
|
|
|
|
KASSERT(sc->sc_width == 2, ("sc_width %d", sc->sc_width));
|
|
|
|
|
2009-02-07 05:03:25 +00:00
|
|
|
#ifdef CFI_ARMEDANDDANGEROUS
|
2009-02-05 18:12:07 +00:00
|
|
|
for (i = 7; i >= 4; i--, id >>= 16) {
|
|
|
|
intr = intr_disable();
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
start = sbinuptime();
|
2009-02-05 18:12:07 +00:00
|
|
|
cfi_write(sc, 0, CFI_INTEL_PP_SETUP);
|
|
|
|
cfi_put16(sc, CFI_INTEL_PR(i), id&0xffff);
|
|
|
|
intr_restore(intr);
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
error = cfi_wait_ready(sc, CFI_BCS_READ_STATUS, start,
|
|
|
|
CFI_TIMEOUT_WRITE);
|
2009-02-05 18:12:07 +00:00
|
|
|
if (error)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
|
|
|
return error;
|
2009-02-07 05:03:25 +00:00
|
|
|
#else
|
|
|
|
device_printf(sc->sc_dev, "%s: OEM PR not set, "
|
|
|
|
"CFI_ARMEDANDDANGEROUS not configured\n", __func__);
|
|
|
|
return ENXIO;
|
|
|
|
#endif
|
2009-02-05 18:12:07 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read the contents of the Protection Lock Register.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
cfi_intel_get_plr(struct cfi_softc *sc, uint32_t *plr)
|
|
|
|
{
|
|
|
|
if (sc->sc_cmdset != CFI_VEND_INTEL_ECS)
|
|
|
|
return EOPNOTSUPP;
|
|
|
|
KASSERT(sc->sc_width == 2, ("sc_width %d", sc->sc_width));
|
|
|
|
|
|
|
|
cfi_write(sc, 0, CFI_INTEL_READ_ID);
|
|
|
|
*plr = cfi_get16(sc, CFI_INTEL_PLR);
|
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Write the Protection Lock Register to lock down the
|
|
|
|
* user-settable segment of the Protection Register.
|
|
|
|
* NOTE: this operation is not reversible.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
cfi_intel_set_plr(struct cfi_softc *sc)
|
|
|
|
{
|
|
|
|
#ifdef CFI_ARMEDANDDANGEROUS
|
|
|
|
register_t intr;
|
|
|
|
int error;
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
sbintime_t start;
|
2009-02-07 05:32:19 +00:00
|
|
|
#endif
|
2009-02-05 18:12:07 +00:00
|
|
|
if (sc->sc_cmdset != CFI_VEND_INTEL_ECS)
|
|
|
|
return EOPNOTSUPP;
|
|
|
|
KASSERT(sc->sc_width == 2, ("sc_width %d", sc->sc_width));
|
|
|
|
|
|
|
|
#ifdef CFI_ARMEDANDDANGEROUS
|
|
|
|
/* worthy of console msg */
|
|
|
|
device_printf(sc->sc_dev, "set PLR\n");
|
|
|
|
intr = intr_disable();
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
binuptime(&start);
|
2009-02-05 18:12:07 +00:00
|
|
|
cfi_write(sc, 0, CFI_INTEL_PP_SETUP);
|
|
|
|
cfi_put16(sc, CFI_INTEL_PLR, 0xFFFD);
|
|
|
|
intr_restore(intr);
|
MFP4 217312, 222008, 222052, 222053, 222673, 231484, 231491, 231565, 570643
Rework the timeout code to use actual time rather than a DELAY() loop and
to use both typical and maximum to allow logging of timeout failures.
Also correct the erase timeout, it is specified in milliseconds not
microseconds like the other timeouts. Do not invoke DELAY() between
status queries as this adds significant latency which in turn reduced
write performance substantially.
Sanity check timeout values from the hardware.
Implement support for buffered writes (only enabled on Intel/Sharp parts
for now). This yields an order of magnitude speedup on the 64MB Intel
StrataFlash parts we use.
When making a copy of the block to modify, also keep a clean copy around
until we are ready to commit the block and use it to avoid unnecessary
erases. In the non-buffer write case, also use it to avoid
unnecessary writes when the block has not been erased. This yields a
significant speedup when doing things like zeroing a block.
Sponsored by: DARPA, AFRL
Reviewed by: imp (previous version)
2013-09-04 17:19:21 +00:00
|
|
|
error = cfi_wait_ready(sc, CFI_BCS_READ_STATUS, start,
|
|
|
|
CFI_TIMEOUT_WRITE);
|
2009-02-05 18:12:07 +00:00
|
|
|
cfi_write(sc, 0, CFI_BCS_READ_ARRAY);
|
2009-02-07 05:32:19 +00:00
|
|
|
return error;
|
2009-02-05 18:12:07 +00:00
|
|
|
#else
|
|
|
|
device_printf(sc->sc_dev, "%s: PLR not set, "
|
|
|
|
"CFI_ARMEDANDDANGEROUS not configured\n", __func__);
|
2009-02-07 05:32:19 +00:00
|
|
|
return ENXIO;
|
2009-02-05 18:12:07 +00:00
|
|
|
#endif
|
|
|
|
}
|
|
|
|
#endif /* CFI_SUPPORT_STRATAFLASH */
|