Accessing PCI-E resources on 460EX
Hi,
I'm currently struggling with some application that tries to access PCI-E memory by mmaping address found in /sys/bus/pci/devices/<device>/resource
mmap() fails with "invalid argument" error, apparently because on 460EX PCI addresses are 36-bit.
What is the correct way of accessing PCI-E from userspace on this platform.
Thanks a lot.
Felix Radensky
Embedded Solutions Ltd.
http://www.embedded-sol.com
[email protected] <> said on :
Quoted 9 lines
I'm currently struggling with some application that tries to access PCI-E memory by mmaping address found in /sys/bus/pci/devices/<device>/resource
=20mmap()fails with "invalid argument" error, apparently because on 460EX PCI addresses are 36-bit.
=20
What is the correct way of accessing PCI-E from userspace on this platform.=20
What works for me is using the mmap64 as in the following example to map the physical address, fd is the file descriptor for /dev/mem
off64_t offset =3D static_cast<off64_t>(your physical address); void * const p =3D mmap64(0,256,PROT_WRITE|PROT_READ,MAP_SHARED,fd,offset);
The only thing to note is that the address you pass should be aligned=20 to the page boundry even though some devices are not.
Quoted 10 lines · Ayman El-Khashab
What works for me is using the mmap64 as in the following example to map the physical address, fd is the file descriptor for
/dev/mem
off64_toffset = static_cast<off64_t>(your physical address); void * const p =mmap64(0,256,PROT_WRITE|PROT_READ,MAP_SHARED,fd,offset);The only thing to note is that the address you pass should be aligned to the page boundry even though some devices are not.
Thanks, Ayman. I've tried that, but I'm still getting the same error. My physical address is 0xe80000000, length is 0x100000.
Felix.
Did you try mmap2. It probable is not in glibc so you will have to call it using syscall.
#include <sys/syscall.h>
unsigned *pdata = (unsigned *)syscall(__NR_mmap2,0,0x100000,PROT_READ| PROT_WRITE,MAP_SHARED,fd, 0xe80000 );
The address is in page offsets that why the address is 0xe80000.
Quoted 21 lines · Felix Radensky
Quoted 11 lines · Ayman El-Khashab
What works for me is using the mmap64 as in the following example to map the physical address, fd is the file descriptor for
/dev/mem
off64_toffset = static_cast<off64_t>(your physical address); void * const p =mmap64(0,256,PROT_WRITE|PROT_READ,MAP_SHARED,fd,offset);The only thing to note is that the address you pass should be aligned to the page boundry even though some devices are not.
Thanks, Ayman. I've tried that, but I'm still getting the same error. My physical address is
0xe80000000, length is0x100000.Felix.
_______________________________________________
Linuxppc-embedded mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-embedded
Quoted 11 lines · Felix Radensky
Hi,
I'm currently struggling with some application that tries to access PCI-E memory by mmaping address found in /sys/bus/pci/devices/<device>/resource
mmap()fails with "invalid argument" error, apparently because on 460EX PCI addresses are 36-bit.What is the correct way of accessing PCI-E from userspace on this platform.
I fixed a lot of these problems recently. Can you try again with 2.6.28-latest-rc ?
Cheers,
Ben.
Quoted 11 lines · Benjamin Herrenschmidt
Quoted 9 lines · Felix Radensky
Hi,
I'm currently struggling with some application that tries to access PCI-E memory by mmaping address found in /sys/bus/pci/devices/<device>/resource
mmap()fails with "invalid argument" error, apparently because on 460EX PCI addresses are 36-bit.
Quoted 7 lines
Quoted 3 lines
What is the correct way of accessing PCI-E from userspace on this platform.
I fixed a lot of these problems recently. Can you try again with 2.6.28-latest-rc ?
Thanks a lot for this input.
The problem is that I have a custom board with quite a few kernel modifications, and I'm afraid it can take me some time to move from 2.6.26 to 2.6.28. Any chance I can backport you modifications to 2.6.26 ? Have you modified other files except arch/powerpc/sysdev/ppc4xx_pci.c to fix these problems ?
Quoted 8 lines
Cheers,
Ben._______________________________________________
Linuxppc-embedded mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-embedded
Quoted 6 lines · Felix Radensky
and I'm afraid it can take me some time to move from 2.6.26 to 2.6.28. Any chance I can backport you modifications to 2.6.26 ? Have you modified
other files exceptarch/powerpc/sysdev/ppc4xx_pci.cto fix these problems ?
Yes, quite a few changes in the PCI layer etc...
Board support changes shouldn't be too hard to port over tho, it depends how you did them.
Cheers,
Ben.
Quoted 10 lines · Benjamin Herrenschmidt
Quoted 7 lines · Felix Radensky
and I'm afraid it can take me some time to move from 2.6.26 to 2.6.28. Any chance I can backport you modifications to 2.6.26 ? Have you modified
other files exceptarch/powerpc/sysdev/ppc4xx_pci.cto fix these problems ?
Quoted 5 lines
Yes, quite a few changes in the PCI layer etc...
Board support changes shouldn't be too hard to port over tho, it depends how you did them.
I've managed to run 2.6.28-rc6 on this platform, it was not that difficult after all. I only had to remove PCIX and PCIE0 from Canyonlands DTS, as they are not used on this board. However this did not solve my problem. Below are relevant bits from dmesg
PCIE1: Checking link...
PCIE1: Device detected, waiting for link...
PCIE1: link is up !
PCI host bridge /plb/pciex@d20000000 (primary) ranges:
MEM 0x0000000e80000000..0x0000000effffffff -> 0x0000000080000000
IO 0x0000000f80010000..0x0000000f8001ffff -> 0x0000000000000000
4xx PCI DMA offset set to 0x00000000
PCIE1: successfully set as root-complex
PCI: Probing PCI hardware
PCI: Hiding 4xx host bridge resources 0000:80:00.0
pci 0000:81:00.0: disabling ASPM on pre-1.1 PCIe device. You can enable it with 'pcie_aspm=force' pci 0000:80:00.0: PCI bridge, secondary bus 0000:81
pci 0000:80:00.0: IO window: disabled pci 0000:80:00.0: MEM window: 0x80000000-0x800fffff pci 0000:80:00.0: PREFETCH window: disabled
bus: 80 index 0 io port: [0x00-0xffff]
bus: 80 index 1 mmio: [0xe80000000-0xeffffffff]
bus: 81 index 0 mmio: [0x0-0xfff]
bus: 81 index 1 mmio: [0xe80000000-0xe800fffff]
bus: 81 index 2 mmio: [0x0-0x0]
bus: 81 index 3 mmio: [0x0-0x0]
The application code that mmaps PCI memory is attached. The attempt to access /sys/bus/pci/devices/0000:81:00.0/resource0 results in I/O error. The device in question is Mellanox Infiniband switch. I will gladly test any other ideas you may have.
Felix.
Quoted 9 lines
Cheers,
Ben._______________________________________________
Linuxppc-embedded mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-embedded
/*
*
* mtcr.h - Mellanox Software tools (mst) driver definitions *
* Author: Michael S. Tsirkin <[email protected]>
*
* Copyright (c) 2006 Mellanox Technologies Ltd. All rights reserved. *
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or * without modification, are permitted provided that the following * conditions are met: * * - Redistributions of source code must retain the above * copyright notice, this list of conditions and the following * disclaimer. * * - 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.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#ifndef _MST_H
#define _MST_H
//use memory mapped /dev/mem for access
#define CONFIG_ENABLE_MMAP 1
//mmap /dev/mem for memory access (does not work on sparc) #define CONFIG_USE_DEV_MEM 1
//use pci configuration cycles for access
#define CONFIG_ENABLE_PCICONF 1
#if CONFIG_ENABLE_PCICONF && CONFIG_ENABLE_MMAP
/* For strerror_r */
#define _XOPEN_SOURCE 600
#elif CONFIG_ENABLE_PCICONF
#define _XOPEN_SOURCE 500
#endif
#if CONFIG_ENABLE_MMAP
#define _FILE_OFFSET_BITS 64
#endif
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <netinet/in.h>
#include <endian.h>
#include <byteswap.h>
#include <errno.h>
#include <fcntl.h>
#include <string.h>
#include <stdlib.h>
#include <libgen.h>
#if CONFIG_ENABLE_MMAP
#include <sys/mman.h>
#include <sys/pci.h>
#include <sys/ioctl.h>
#include <sys/types.h>
#endif
#ifndef __be32_to_cpu
#define __be32_to_cpu(x) ntohl(x)
#endif
#ifndef __cpu_to_be32
#define __cpu_to_be32(x) htonl(x)
#endif
#if __BYTE_ORDER == __LITTLE_ENDIAN
#ifndef __cpu_to_le32
#define __cpu_to_le32(x) (x)
#endif
#ifndef __le32_to_cpu
#define __le32_to_cpu(x) (x)
#endif
#elif __BYTE_ORDER == __BIG_ENDIAN
#ifndef __cpu_to_le32
#define __cpu_to_le32(x) bswap_32(x)
#endif
#ifndef __le32_to_cpu
#define __le32_to_cpu(x) bswap_32(x)
#endif
#else
#ifndef __cpu_to_le32
#define __cpu_to_le32(x) bswap_32(__cpu_to_be32(x)) #endif
#ifndef __le32_to_cpu
#define __le32_to_cpu(x) __be32_to_cpu(bswap_32(x)) #endif
#endif
#ifdef __cplusplus
extern "C" {
#endif
/* All fields in the following structure are not supposed to be used */ /* or modified by user programs. */
typedef struct mfile_t {
int fd;
void *ptr;
int connectx_flush; /* For ConnectX A0 */
int need_flush; /* For ConnectX A0 */} mfile;
static void mtcr_connectx_flush(void *ptr)
{
u_int32_t value;
*((u_int32_t *)((char *)ptr + 0xf0380)) = 0x0;
do {
asm volatile ("":::"memory");
value = __be32_to_cpu(*((u_int32_t *)((char *)ptr + 0xf0380)));
} while(value);
}/*
* Read 4 bytes, return number of succ. read bytes or -1 on failure */
int mread4(mfile *mf, unsigned int offset, u_int32_t *value) #ifdef MTCR_EXPORT
;
#else
{
#if CONFIG_ENABLE_MMAP
if (mf->ptr) {
if (mf->need_flush) {
mtcr_connectx_flush(mf->ptr);
mf->need_flush = 0;
}
*value = __be32_to_cpu(*((u_int32_t *)((char *)mf->ptr + offset)));
return 4;
}
#endif
#if CONFIG_ENABLE_PCICONF
{
int rc;
offset = __cpu_to_le32(offset);
rc=pwrite(mf->fd, &offset, 4, 22*4);
if (rc < 0) {
perror("write offset");
return rc;
}
if (rc != 4)
return 0;
rc=pread(mf->fd, value, 4, 23*4);
if (rc < 0) {
perror("read value");
return rc;
}
*value = __le32_to_cpu(*value);
return rc;
}
#else
return 0;#endif
}
#endif
/*
* Write 4 bytes, return number of succ. written bytes or -1 on failure */
int mwrite4(mfile *mf, unsigned int offset, u_int32_t value) #ifdef MTCR_EXPORT
;
#else
{
#if CONFIG_ENABLE_MMAP
if (mf->ptr) {
*((u_int32_t *)((char *)mf->ptr + offset)) = __cpu_to_be32(value);
mf->need_flush = mf->connectx_flush;
return 4;
}
#endif
#if CONFIG_ENABLE_PCICONF
{
int rc;
offset = __cpu_to_le32(offset);
rc = pwrite(mf->fd, &offset, 4, 22*4);
if (rc < 0) {
perror("write offset");
return rc;
}
if (rc != 4)
return 0;
value = __cpu_to_le32(value);
rc = pwrite(mf->fd, &value, 4, 23*4);
if (rc < 0) {
perror("write value");
return rc;
}
return rc; }
#else
return 0;
#endif
}
#endif
#ifndef MTCR_EXPORT
static intmread_chunk_as_multi_mread4(mfile *mf, unsigned int offset, void *data, int length)
{
int i;
if (length % 4) {
return EINVAL;
}
for (i = 0; i < length ; i += 4) {
u_int32_t value;
if (mread4(mf, offset + i, &value) != 4) {
return -1;
}
memcpy((char*)data + i , &value,4);
}
return length;
}static intmwrite_chunk_as_multi_mwrite4(mfile *mf, unsigned int offset, void *data, int length)
{
int i;
if (length % 4) {
return EINVAL;
}
for (i = 0; i < length ; i += 4) {
u_int32_t value;
memcpy(&value, (char*)data + i ,4);
if (mwrite4(mf, offset + i, value) != 4) {
return -1;
}
}
return length;
}enum mtcr_access_method {
MTCR_ACCESS_ERROR = 0x0,
MTCR_ACCESS_MEMORY = 0x1,
MTCR_ACCESS_CONFIG = 0x2,
};static
int mtcr_check_signature(mfile *mf)
{
unsigned signature;
int rc;
rc = mread4(mf, 0xF0014, &signature);
if (rc != 4) {
if (!errno)
errno = EIO;
return -1;
}
switch (signature & 0xffff) {
case 0x190 : /* 400 */
if (signature == 0xa00190 && mf->ptr) {
mf->connectx_flush = 1;
mtcr_connectx_flush(mf->ptr);
}
case 0x5a44: /* 23108 */
case 0x6278: /* 25208 */
case 0x5e8c: /* 24204 */
case 0x6274: /* 25204 */
return 0;
default:
errno = ENOTTY;
return -1;
}
}#if CONFIG_ENABLE_MMAP
/*
* The PCI interface treats multi-function devices as independent
* devices. The slot/function address of each device is encoded
* in a single byte as follows:
*
* 7:3 = slot
* 2:0 = function
*/
#define PCI_DEVFN(slot,func) ((((slot) & 0x1f) << 3) | ((func) & 0x07))
#define PCI_SLOT(devfn) (((devfn) >> 3) & 0x1f)
#define PCI_FUNC(devfn) ((devfn) & 0x07)
static
unsigned long long mtcr_procfs_get_offset(unsigned my_bus, unsigned my_dev,
unsigned my_func)
{
FILE* f;
unsigned irq;
unsigned long long base_addr[6], rom_base_addr, size[6], rom_size;
unsigned bus, dev, func;
unsigned vendor_id;
unsigned device_id;
unsigned int cnt;
unsigned long long offset = (unsigned long long)-1;
char buf[4048];
f = fopen("/proc/bus/pci/devices", "r");
if (!f) return offset;
for(;;) if (fgets(buf, sizeof(buf) - 1, f)) {
unsigned dfn, vend;
cnt = sscanf(buf,
"%x %x %x %llx %llx %llx %llx %llx %llx "
"%llx %llx %llx %llx %llx %llx %llx %llx",
&dfn,
&vend,
&irq,
&base_addr[0],
&base_addr[1],
&base_addr[2],
&base_addr[3],
&base_addr[4],
&base_addr[5],
&rom_base_addr,
&size[0],
&size[1],
&size[2],
&size[3],
&size[4],
&size[5],
&rom_size);
if (cnt != 9 && cnt != 10 && cnt != 17)
{
fprintf(stderr,"proc: parse error (read only %d items)\n", cnt);
fprintf(stderr,"the offending line in " "/proc/bus/pci/devices" " is "
"\"%.*s\"\n", (int)sizeof(buf), buf);
goto error;
}
bus = dfn >> 8U;
dev = PCI_SLOT(dfn & 0xff);
func = PCI_FUNC(dfn & 0xff);
vendor_id = vend >> 16U;
device_id = vend & 0xffff;
if (bus == my_bus && dev == my_dev && func == my_func)
break;
}
else
goto error;
if (cnt != 17 || size[1] != 0 || size[0] != 0x100000) {
if (0) fprintf(stderr,"proc: unexpected region size values: "
"cnt=%d, size[0]=%#llx, size[1]=%#llx\n",
cnt,size[0],size[1]);
if (0) fprintf(stderr,"the offending line in " "/proc/bus/pci/devices"
" is \"%.*s\"\n", (int)sizeof(buf), buf);
goto error;
}
offset = ((unsigned long long)(base_addr[1]) << 32) +
((unsigned long long)(base_addr[0]) & ~(unsigned long long)(0xfffff));
fclose(f);
return offset;error:
fclose(f); errno = ENXIO; return offset; }
static
unsigned long long mtcr_sysfs_get_offset(unsigned domain, unsigned bus,
unsigned dev, unsigned func)
{
unsigned long long start, end, type;
unsigned long long offset = (unsigned long long)-1;
FILE *f;
int cnt;
char mbuf[] = "/sys/bus/pci/devices/XXXX:XX:XX.X/resource";
sprintf(mbuf, "/sys/bus/pci/devices/%4.4x:%2.2x:%2.2x.%1.1x/resource",
domain, bus, dev, func);
f = fopen(mbuf, "r");
if (!f)
return offset;
cnt = fscanf(f, "0x%llx 0x%llx 0x%llx", &start, &end, &type);
if (cnt != 3 || end != start + 0x100000 - 1) {
if (0) fprintf(stderr,"proc: unexpected region size values: "
"cnt=%d, start=%#llx, end=%#llx\n",
cnt, start, end);
goto error;
}
fclose(f);
return start;error:
fclose(f); errno = ENOENT; return offset; }
static
int mtcr_mmap(mfile *mf, const char *name, off_t off, int ioctl_needed)
{
int err;
mf->fd = open(name, O_RDWR | O_SYNC);
if (mf->fd < 0)
return -1;
if (ioctl_needed && ioctl(mf->fd, PCIIOC_MMAP_IS_MEM) < 0) {
err = errno;
close(mf->fd);
errno = err;
return -1;
}
mf->ptr = mmap(NULL, 0x100000, PROT_READ | PROT_WRITE,
MAP_SHARED, mf->fd, off);
if (!mf->ptr || mf->ptr == MAP_FAILED) {
err = errno;
close(mf->fd);
errno = err;
return -1;
}
if (mtcr_check_signature(mf)) {
munmap(mf->ptr, 0x10000);
close(mf->fd);
errno = EIO;
return -1;
}
return 0;
}#else
static
unsigned long long mtcr_procfs_get_offset(unsigned my_bus, unsigned my_dev,
unsigned my_func)
{
return -1;
}static
unsigned long long mtcr_sysfs_get_offset(unsigned domain, unsigned bus,
unsigned dev, unsigned func)
{
return -1;
}static
int mtcr_mmap(mfile *mf, const char *name, off_t off, int ioctl_needed)
{
return -1;
}#endif
#if CONFIG_ENABLE_PCICONF
static
int mtcr_open_config(mfile *mf, const char *name)
{
unsigned signature;
int rc, err;
mf->fd = open(name, O_RDWR | O_SYNC);
if (mf->fd < 0)
return -1;
/* Kernels before 2.6.12 carry the high bit in each byte
* on <device>/config writes, overriding higher bits.
* Make sure the high bit is set in some signature bytes,
* to catch this. */
/* Do this test before mtcr_check_signature,
to avoid system failure on access to an illegal address. */
signature = 0xfafbfcfd;
rc = pwrite(mf->fd, &signature, 4, 22*4);
if (rc != 4) {
err = errno;
close(mf->fd);
errno = err;
return -1;
}
rc = pread(mf->fd, &signature, 4, 22*4);
if (rc != 4) {
err = errno;
close(mf->fd);
errno = err;
return -1;
}
if (signature != 0xfafbfcfd) {
close(mf->fd);
errno = EIO;
return -1;
}
if (mtcr_check_signature(mf)) {
close(mf->fd);
errno = EIO;
return -1;
}
return 0;
}
#else
static
int mtcr_open_config(mfile *mf, const char *name)
{
return -1;
}#endif
static
enum mtcr_access_method mtcr_parse_name(const char* name, int *force,
unsigned *domain_p, unsigned *bus_p, unsigned *dev_p, unsigned *func_p)
{
unsigned my_domain = 0;
unsigned my_bus;
unsigned my_dev;
unsigned my_func;
int scnt, r;
char config[] = "/config";
char resource0[] = "/resource0";
char procbuspci[] = "/proc/bus/pci/";
unsigned len = strlen(name);
unsigned tmp;
if (len >= sizeof config && !strcmp(config, name + len + 1 - sizeof config)) {
*force = 1;
return MTCR_ACCESS_CONFIG;
}
if (len >= sizeof resource0 &&
!strcmp(resource0, name + len + 1 - sizeof resource0)) {
*force = 1;
return MTCR_ACCESS_MEMORY;
}
if (!strncmp(name,"/proc/bus/pci/", sizeof procbuspci - 1)) {
*force = 1;
return MTCR_ACCESS_CONFIG;
}
if (sscanf(name, "mthca%x", &tmp) == 1 ||
sscanf(name, "mlx4_%x", &tmp) == 1) {
char mbuf[4048];
char pbuf[4048];
char *base;
r = snprintf(mbuf, sizeof mbuf, "/sys/class/infiniband/%s/device", name);
if (r <= 0 || r >= (int)sizeof mbuf) {
fprintf(stderr,"Unable to print device name %s\n", name);
goto parse_error;
}
r = readlink(mbuf, pbuf, sizeof pbuf - 1);
if (r < 0) {
perror("read link");
fprintf(stderr,"Unable to read link %s\n", mbuf);
return MTCR_ACCESS_ERROR;
}
pbuf[r] = '\0';
base = basename(pbuf);
if (!base)
goto parse_error;
scnt = sscanf(base, "%x:%x:%x.%x",
&my_domain, &my_bus, &my_dev, &my_func);
if (scnt != 4)
goto parse_error;
goto name_parsed;
}
scnt = sscanf(name, "%x:%x.%x", &my_bus, &my_dev, &my_func);
if (scnt == 3)
goto name_parsed;
scnt = sscanf(name, "%x:%x:%x.%x", &my_domain, &my_bus, &my_dev, &my_func);
if (scnt == 4)
goto name_parsed;parse_error:
fprintf(stderr,"Unable to parse device name %s\n", name); errno = EINVAL; return MTCR_ACCESS_ERROR;
name_parsed:
*domain_p = my_domain; *bus_p = my_bus; *dev_p = my_dev; *func_p = my_func; *force = 0; return MTCR_ACCESS_MEMORY; }
#endif
int mread4_block (mfile *mf, unsigned int offset, u_int32_t* data, int byte_len)
#ifdef MTCR_EXPORT
;
#else
{
return mread_chunk_as_multi_mread4(mf, offset, data, byte_len);
}#endif
int mwrite4_block (mfile *mf, unsigned int offset, u_int32_t* data, int byte_len)
#ifdef MTCR_EXPORT
;
#else
{
return mwrite_chunk_as_multi_mwrite4(mf, offset, data, byte_len);
}#endif
int msw_reset(mfile *mf)
#ifdef MTCR_EXPORT
;
#else
{
mf = NULL; /* Warning */
return -1;
}#endif
/*
* Open Mellanox Software tools (mst) driver. Device type==TAVOR
* Return valid mfile ptr or 0 on failure
*/
mfile *mopen(const char *name)
#ifdef MTCR_EXPORT
;
#else
{
mfile *mf;
off_t offset;
unsigned domain, bus, dev, func;
enum mtcr_access_method access;
int force;
char rbuf[] = "/sys/bus/pci/devices/XXXX:XX:XX.X/resource0";
char cbuf[] = "/sys/bus/pci/devices/XXXX:XX:XX.X/config";
char pdbuf[] = "/proc/bus/pci/XXXX:XX/XX.X";
char pbuf[] = "/proc/bus/pci/XX/XX.X";
char errbuf[4048]="";
int err;
mf = (mfile *)malloc(sizeof(mfile));
if (!mf)
return NULL;
mf->ptr = NULL;
mf->fd = -1;
mf->connectx_flush = mf->need_flush = 0;
access = mtcr_parse_name(name, &force, &domain, &bus, &dev, &func);
if (access == MTCR_ACCESS_ERROR)
goto open_failed;
if (force) {
if (access == MTCR_ACCESS_CONFIG) {
if (!mtcr_open_config(mf, name))
return mf;
} else {
if (!mtcr_mmap(mf, name, 0, 0))
return mf;
}
goto open_failed;
}
if (access == MTCR_ACCESS_CONFIG)
goto access_config_forced;
sprintf(rbuf, "/sys/bus/pci/devices/%4.4x:%2.2x:%2.2x.%1.1x/resource0",
domain, bus, dev, func);
if (!mtcr_mmap(mf, rbuf, 0, 0))
return mf;
/* Following access methods need the resource BAR */
offset = mtcr_sysfs_get_offset(domain, bus, dev, func);
if (offset == -1 && !domain)
offset = mtcr_procfs_get_offset(bus, dev, func);
if (offset == -1)
goto access_config;
sprintf(pdbuf, "/proc/bus/pci/%4.4x:%2.2x/%2.2x.%1.1x",
domain, bus, dev, func);
if (!mtcr_mmap(mf, pdbuf, offset, 1))
return mf;
if (!domain) {
sprintf(pbuf, "/proc/bus/pci/%2.2x/%2.2x.%1.1x",
bus, dev, func);
if (!mtcr_mmap(mf, pbuf, offset, 1))
return mf;
}#if CONFIG_USE_DEV_MEM /* Non-portable, but helps some systems */ if (!mtcr_mmap(mf, "/dev/mem", offset, 0)) return mf;
#endif
access_config:
#if CONFIG_ENABLE_PCICONF && CONFIG_ENABLE_PCICONF strerror_r(errno, errbuf, sizeof errbuf); fprintf(stderr, "Warning: memory access to device %s failed: %s.\n" "Warning: Fallback on IO: much slower, and unsafe if device in use.\n", name, errbuf);
#endif
access_config_forced:
sprintf(cbuf, "/sys/bus/pci/devices/%4.4x:%2.2x:%2.2x.%1.1x/config",
domain, bus, dev, func);
if (!mtcr_open_config(mf, cbuf))
return mf;
sprintf(pdbuf, "/proc/bus/pci/devices/%4.4x:%2.2x/%2.2x.%1.1x",
domain, bus, dev, func);
if (!mtcr_open_config(mf, pdbuf))
return mf;
if (!domain) {
sprintf(pbuf, "/proc/bus/pci/devices/%2.2x/%2.2x.%1.1x",
bus, dev, func);
if (!mtcr_open_config(mf, pdbuf))
return mf;
}open_failed:
err = errno;
free(mf);
errno = err;
return NULL;
}#endif
/*
* Close Mellanox driver
* req. descriptor
*/
int mclose(mfile *mf)
#ifdef MTCR_EXPORT
;
#else
{
#if CONFIG_ENABLE_MMAP
if (mf->ptr)
munmap(mf->ptr,0x10000);
#endif
close(mf->fd);
free(mf);
return 0;
}
#endif
#ifdef __cplusplus
}
#endif
#define DEV_MST_EXAMPLE1 "mthca0"
#define DEV_MST_EXAMPLE2 "03:00.0"
#endif
Hi, Ben
Sorry for confusion. With 2.6.28-rc6 mmap() of resource0 works fine. The application still doesn't work, but that's because it's not reading what it expects from resource file. I guess it's no longer a kernel problem.
Thanks a lot for your help.
Felix.
Quoted 66 lines · Felix Radensky
Quoted 10 lines · Benjamin Herrenschmidt
Quoted 7 lines · Felix Radensky
and I'm afraid it can take me some time to move from 2.6.26 to 2.6.28. Any chance I can backport you modifications to 2.6.26 ? Have you modified
other files exceptarch/powerpc/sysdev/ppc4xx_pci.cto fix these problems ?Quoted 5 lines
Yes, quite a few changes in the PCI layer etc...
Board support changes shouldn't be too hard to port over tho, it depends how you did them.
I've managed to run 2.6.28-rc6 on this platform, it was not that difficult after all. I only had to remove PCIX and PCIE0 from Canyonlands DTS, as they are not used on this board. However this did not solve my problem. Below are relevant bits from dmesg
PCIE1: Checking link...
PCIE1: Device detected, waiting for link...
PCIE1: link is up !
PCI host bridge /plb/pciex@d20000000 (primary) ranges:
MEM 0x0000000e80000000..0x0000000effffffff ->0x0000000080000000
IO 0x0000000f80010000..0x0000000f8001ffff ->0x0000000000000000
4xx PCI DMA offset set to0x00000000
PCIE1: successfully set as root-complex
PCI: Probing PCI hardware
PCI: Hiding 4xx host bridge resources 0000:80:00.0
pci 0000:81:00.0: disabling ASPM on pre-1.1 PCIe device. You can enable it with 'pcie_aspm=force' pci 0000:80:00.0: PCI bridge, secondary bus 0000:81pci 0000:80:00.0: IO window: disabled pci 0000:80:00.0: MEM window: 0x80000000-0x800fffff pci 0000:80:00.0: PREFETCH window: disabledbus: 80 index 0 io port: [0x00-0xffff]
bus: 80 index 1 mmio: [0xe80000000-0xeffffffff]
bus: 81 index 0 mmio: [0x0-0xfff]
bus: 81 index 1 mmio: [0xe80000000-0xe800fffff]
bus: 81 index 2 mmio: [0x0-0x0]
bus: 81 index 3 mmio: [0x0-0x0]The application code that mmaps PCI memory is attached. The attempt to access /sys/bus/pci/devices/0000:81:00.0/resource0 results in I/O error. The device in question is Mellanox Infiniband switch. I will gladly test any other ideas you may have.
Felix.
Quoted 9 lines
Cheers,
Ben._______________________________________________
Linuxppc-embedded mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-embedded------------------------------------------------------------------------
_______________________________________________
Linuxppc-embedded mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-embedded