OpenCloudOS-Kernel/drivers/edac/i82875p_edac.c

601 lines
15 KiB
C
Raw Normal View History

/*
* Intel D82875P Memory Controller kernel module
* (C) 2003 Linux Networx (http://lnxi.com)
* This file may be distributed under the terms of the
* GNU General Public License.
*
* Written by Thayne Harbaugh
* Contributors:
* Wang Zhenyu at intel.com
*
* $Id: edac_i82875p.c,v 1.5.2.11 2005/10/05 00:43:44 dsp_llnl Exp $
*
* Note: E7210 appears same as D82875P - zhenyu.z.wang at intel.com
*/
#include <linux/module.h>
#include <linux/init.h>
#include <linux/pci.h>
#include <linux/pci_ids.h>
#include <linux/edac.h>
#include "edac_module.h"
#define EDAC_MOD_STR "i82875p_edac"
#define i82875p_printk(level, fmt, arg...) \
edac_printk(level, "i82875p", fmt, ##arg)
#define i82875p_mc_printk(mci, level, fmt, arg...) \
edac_mc_chipset_printk(mci, level, "i82875p", fmt, ##arg)
#ifndef PCI_DEVICE_ID_INTEL_82875_0
#define PCI_DEVICE_ID_INTEL_82875_0 0x2578
#endif /* PCI_DEVICE_ID_INTEL_82875_0 */
#ifndef PCI_DEVICE_ID_INTEL_82875_6
#define PCI_DEVICE_ID_INTEL_82875_6 0x257e
#endif /* PCI_DEVICE_ID_INTEL_82875_6 */
/* four csrows in dual channel, eight in single channel */
#define I82875P_NR_DIMMS 8
#define I82875P_NR_CSROWS(nr_chans) (I82875P_NR_DIMMS / (nr_chans))
/* Intel 82875p register addresses - device 0 function 0 - DRAM Controller */
#define I82875P_EAP 0x58 /* Error Address Pointer (32b)
*
* 31:12 block address
* 11:0 reserved
*/
#define I82875P_DERRSYN 0x5c /* DRAM Error Syndrome (8b)
*
* 7:0 DRAM ECC Syndrome
*/
#define I82875P_DES 0x5d /* DRAM Error Status (8b)
*
* 7:1 reserved
* 0 Error channel 0/1
*/
#define I82875P_ERRSTS 0xc8 /* Error Status Register (16b)
*
* 15:10 reserved
* 9 non-DRAM lock error (ndlock)
* 8 Sftwr Generated SMI
* 7 ECC UE
* 6 reserved
* 5 MCH detects unimplemented cycle
* 4 AGP access outside GA
* 3 Invalid AGP access
* 2 Invalid GA translation table
* 1 Unsupported AGP command
* 0 ECC CE
*/
#define I82875P_ERRCMD 0xca /* Error Command (16b)
*
* 15:10 reserved
* 9 SERR on non-DRAM lock
* 8 SERR on ECC UE
* 7 SERR on ECC CE
* 6 target abort on high exception
* 5 detect unimplemented cyc
* 4 AGP access outside of GA
* 3 SERR on invalid AGP access
* 2 invalid translation table
* 1 SERR on unsupported AGP command
* 0 reserved
*/
/* Intel 82875p register addresses - device 6 function 0 - DRAM Controller */
#define I82875P_PCICMD6 0x04 /* PCI Command Register (16b)
*
* 15:10 reserved
* 9 fast back-to-back - ro 0
* 8 SERR enable - ro 0
* 7 addr/data stepping - ro 0
* 6 parity err enable - ro 0
* 5 VGA palette snoop - ro 0
* 4 mem wr & invalidate - ro 0
* 3 special cycle - ro 0
* 2 bus master - ro 0
* 1 mem access dev6 - 0(dis),1(en)
* 0 IO access dev3 - 0(dis),1(en)
*/
#define I82875P_BAR6 0x10 /* Mem Delays Base ADDR Reg (32b)
*
* 31:12 mem base addr [31:12]
* 11:4 address mask - ro 0
* 3 prefetchable - ro 0(non),1(pre)
* 2:1 mem type - ro 0
* 0 mem space - ro 0
*/
/* Intel 82875p MMIO register space - device 0 function 0 - MMR space */
#define I82875P_DRB_SHIFT 26 /* 64MiB grain */
#define I82875P_DRB 0x00 /* DRAM Row Boundary (8b x 8)
*
* 7 reserved
* 6:0 64MiB row boundary addr
*/
#define I82875P_DRA 0x10 /* DRAM Row Attribute (4b x 8)
*
* 7 reserved
* 6:4 row attr row 1
* 3 reserved
* 2:0 row attr row 0
*
* 000 = 4KiB
* 001 = 8KiB
* 010 = 16KiB
* 011 = 32KiB
*/
#define I82875P_DRC 0x68 /* DRAM Controller Mode (32b)
*
* 31:30 reserved
* 29 init complete
* 28:23 reserved
* 22:21 nr chan 00=1,01=2
* 20 reserved
* 19:18 Data Integ Mode 00=none,01=ecc
* 17:11 reserved
* 10:8 refresh mode
* 7 reserved
* 6:4 mode select
* 3:2 reserved
* 1:0 DRAM type 01=DDR
*/
enum i82875p_chips {
I82875P = 0,
};
struct i82875p_pvt {
struct pci_dev *ovrfl_pdev;
void __iomem *ovrfl_window;
};
struct i82875p_dev_info {
const char *ctl_name;
};
struct i82875p_error_info {
u16 errsts;
u32 eap;
u8 des;
u8 derrsyn;
u16 errsts2;
};
static const struct i82875p_dev_info i82875p_devs[] = {
[I82875P] = {
.ctl_name = "i82875p"},
};
static struct pci_dev *mci_pdev; /* init dev: in case that AGP code has
* already registered driver
*/
static struct edac_pci_ctl_info *i82875p_pci;
static void i82875p_get_error_info(struct mem_ctl_info *mci,
struct i82875p_error_info *info)
{
struct pci_dev *pdev;
pdev = to_pci_dev(mci->pdev);
/*
* This is a mess because there is no atomic way to read all the
* registers at once and the registers can transition from CE being
* overwritten by UE.
*/
pci_read_config_word(pdev, I82875P_ERRSTS, &info->errsts);
if (!(info->errsts & 0x0081))
return;
pci_read_config_dword(pdev, I82875P_EAP, &info->eap);
pci_read_config_byte(pdev, I82875P_DES, &info->des);
pci_read_config_byte(pdev, I82875P_DERRSYN, &info->derrsyn);
pci_read_config_word(pdev, I82875P_ERRSTS, &info->errsts2);
/*
* If the error is the same then we can for both reads then
* the first set of reads is valid. If there is a change then
* there is a CE no info and the second set of reads is valid
* and should be UE info.
*/
if ((info->errsts ^ info->errsts2) & 0x0081) {
pci_read_config_dword(pdev, I82875P_EAP, &info->eap);
pci_read_config_byte(pdev, I82875P_DES, &info->des);
pci_read_config_byte(pdev, I82875P_DERRSYN, &info->derrsyn);
}
pci_write_bits16(pdev, I82875P_ERRSTS, 0x0081, 0x0081);
}
static int i82875p_process_error_info(struct mem_ctl_info *mci,
struct i82875p_error_info *info,
int handle_errors)
{
int row, multi_chan;
multi_chan = mci->csrows[0]->nr_channels - 1;
if (!(info->errsts & 0x0081))
return 0;
if (!handle_errors)
return 1;
if ((info->errsts ^ info->errsts2) & 0x0081) {
edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0,
-1, -1, -1,
"UE overwrote CE", "");
info->errsts = info->errsts2;
}
info->eap >>= PAGE_SHIFT;
row = edac_mc_find_csrow_by_page(mci, info->eap);
if (info->errsts & 0x0080)
edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1,
info->eap, 0, 0,
row, -1, -1,
"i82875p UE", "");
else
edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1,
info->eap, 0, info->derrsyn,
row, multi_chan ? (info->des & 0x1) : 0,
-1, "i82875p CE", "");
return 1;
}
static void i82875p_check(struct mem_ctl_info *mci)
{
struct i82875p_error_info info;
edac_dbg(1, "MC%d\n", mci->mc_idx);
i82875p_get_error_info(mci, &info);
i82875p_process_error_info(mci, &info, 1);
}
/* Return 0 on success or 1 on failure. */
static int i82875p_setup_overfl_dev(struct pci_dev *pdev,
struct pci_dev **ovrfl_pdev,
void __iomem **ovrfl_window)
{
struct pci_dev *dev;
void __iomem *window;
*ovrfl_pdev = NULL;
*ovrfl_window = NULL;
dev = pci_get_device(PCI_VEND_DEV(INTEL, 82875_6), NULL);
if (dev == NULL) {
/* Intel tells BIOS developers to hide device 6 which
* configures the overflow device access containing
* the DRBs - this is where we expose device 6.
* http://www.x86-secret.com/articles/tweak/pat/patsecrets-2.htm
*/
pci_write_bits8(pdev, 0xf4, 0x2, 0x2);
dev = pci_scan_single_device(pdev->bus, PCI_DEVFN(6, 0));
if (dev == NULL)
return 1;
pci_bus_assign_resources(dev->bus);
pci_bus_add_device(dev);
}
*ovrfl_pdev = dev;
if (pci_enable_device(dev)) {
i82875p_printk(KERN_ERR, "%s(): Failed to enable overflow "
"device\n", __func__);
return 1;
}
if (pci_request_regions(dev, pci_name(dev))) {
#ifdef CORRECT_BIOS
goto fail0;
#endif
}
/* cache is irrelevant for PCI bus reads/writes */
window = pci_ioremap_bar(dev, 0);
if (window == NULL) {
i82875p_printk(KERN_ERR, "%s(): Failed to ioremap bar6\n",
__func__);
goto fail1;
}
*ovrfl_window = window;
return 0;
fail1:
pci_release_regions(dev);
#ifdef CORRECT_BIOS
fail0:
pci_disable_device(dev);
#endif
/* NOTE: the ovrfl proc entry and pci_dev are intentionally left */
return 1;
}
/* Return 1 if dual channel mode is active. Else return 0. */
static inline int dual_channel_active(u32 drc)
{
return (drc >> 21) & 0x1;
}
static void i82875p_init_csrows(struct mem_ctl_info *mci,
struct pci_dev *pdev,
void __iomem * ovrfl_window, u32 drc)
{
struct csrow_info *csrow;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-28 05:38:08 +08:00
struct dimm_info *dimm;
unsigned nr_chans = dual_channel_active(drc) + 1;
unsigned long last_cumul_size;
u8 value;
u32 drc_ddim; /* DRAM Data Integrity Mode 0=none,2=edac */
u32 cumul_size, nr_pages;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-28 05:38:08 +08:00
int index, j;
drc_ddim = (drc >> 18) & 0x1;
last_cumul_size = 0;
/* The dram row boundary (DRB) reg values are boundary address
* for each DRAM row with a granularity of 32 or 64MB (single/dual
* channel operation). DRB regs are cumulative; therefore DRB7 will
* contain the total memory contained in all eight rows.
*/
for (index = 0; index < mci->nr_csrows; index++) {
csrow = mci->csrows[index];
value = readb(ovrfl_window + I82875P_DRB + index);
cumul_size = value << (I82875P_DRB_SHIFT - PAGE_SHIFT);
edac_dbg(3, "(%d) cumul_size 0x%x\n", index, cumul_size);
if (cumul_size == last_cumul_size)
continue; /* not populated */
csrow->first_page = last_cumul_size;
csrow->last_page = cumul_size - 1;
nr_pages = cumul_size - last_cumul_size;
last_cumul_size = cumul_size;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-28 05:38:08 +08:00
for (j = 0; j < nr_chans; j++) {
dimm = csrow->channels[j]->dimm;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-28 05:38:08 +08:00
dimm->nr_pages = nr_pages / nr_chans;
edac: move dimm properties to struct dimm_info On systems based on chip select rows, all channels need to use memories with the same properties, otherwise the memories on channels A and B won't be recognized. However, such assumption is not true for all types of memory controllers. Controllers for FB-DIMM's don't have such requirements. Also, modern Intel controllers seem to be capable of handling such differences. So, we need to get rid of storing the DIMM information into a per-csrow data, storing it, instead at the right place. The first step is to move grain, mtype, dtype and edac_mode to the per-dimm struct. Reviewed-by: Aristeu Rozanski <arozansk@redhat.com> Reviewed-by: Borislav Petkov <borislav.petkov@amd.com> Acked-by: Chris Metcalf <cmetcalf@tilera.com> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Borislav Petkov <borislav.petkov@amd.com> Cc: Mark Gross <mark.gross@intel.com> Cc: Jason Uhlenkott <juhlenko@akamai.com> Cc: Tim Small <tim@buttersideup.com> Cc: Ranganathan Desikan <ravi@jetztechnologies.com> Cc: "Arvind R." <arvino55@gmail.com> Cc: Olof Johansson <olof@lixom.net> Cc: Egor Martovetsky <egor@pasemi.com> Cc: Michal Marek <mmarek@suse.cz> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Joe Perches <joe@perches.com> Cc: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Hitoshi Mitake <h.mitake@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: James Bottomley <James.Bottomley@parallels.com> Cc: "Niklas Söderlund" <niklas.soderlund@ericsson.com> Cc: Shaohui Xie <Shaohui.Xie@freescale.com> Cc: Josh Boyer <jwboyer@gmail.com> Cc: Mike Williams <mike@mikebwilliams.com> Cc: linuxppc-dev@lists.ozlabs.org Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-01-28 05:38:08 +08:00
dimm->grain = 1 << 12; /* I82875P_EAP has 4KiB reolution */
dimm->mtype = MEM_DDR;
dimm->dtype = DEV_UNKNOWN;
dimm->edac_mode = drc_ddim ? EDAC_SECDED : EDAC_NONE;
}
}
}
static int i82875p_probe1(struct pci_dev *pdev, int dev_idx)
{
int rc = -ENODEV;
struct mem_ctl_info *mci;
struct edac_mc_layer layers[2];
struct i82875p_pvt *pvt;
struct pci_dev *ovrfl_pdev;
void __iomem *ovrfl_window;
u32 drc;
u32 nr_chans;
struct i82875p_error_info discard;
edac_dbg(0, "\n");
if (i82875p_setup_overfl_dev(pdev, &ovrfl_pdev, &ovrfl_window))
return -ENODEV;
drc = readl(ovrfl_window + I82875P_DRC);
nr_chans = dual_channel_active(drc) + 1;
layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
layers[0].size = I82875P_NR_CSROWS(nr_chans);
layers[0].is_virt_csrow = true;
layers[1].type = EDAC_MC_LAYER_CHANNEL;
layers[1].size = nr_chans;
layers[1].is_virt_csrow = false;
mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt));
if (!mci) {
rc = -ENOMEM;
goto fail0;
}
edac_dbg(3, "init mci\n");
mci->pdev = &pdev->dev;
mci->mtype_cap = MEM_FLAG_DDR;
mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_SECDED;
mci->edac_cap = EDAC_FLAG_UNKNOWN;
mci->mod_name = EDAC_MOD_STR;
mci->ctl_name = i82875p_devs[dev_idx].ctl_name;
mci->dev_name = pci_name(pdev);
mci->edac_check = i82875p_check;
mci->ctl_page_to_phys = NULL;
edac_dbg(3, "init pvt\n");
pvt = (struct i82875p_pvt *)mci->pvt_info;
pvt->ovrfl_pdev = ovrfl_pdev;
pvt->ovrfl_window = ovrfl_window;
i82875p_init_csrows(mci, pdev, ovrfl_window, drc);
i82875p_get_error_info(mci, &discard); /* clear counters */
/* Here we assume that we will never see multiple instances of this
* type of memory controller. The ID is therefore hardcoded to 0.
*/
if (edac_mc_add_mc(mci)) {
edac_dbg(3, "failed edac_mc_add_mc()\n");
goto fail1;
}
/* allocating generic PCI control info */
i82875p_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
if (!i82875p_pci) {
printk(KERN_WARNING
"%s(): Unable to create PCI control\n",
__func__);
printk(KERN_WARNING
"%s(): PCI error report via EDAC not setup\n",
__func__);
}
/* get this far and it's successful */
edac_dbg(3, "success\n");
return 0;
fail1:
edac_mc_free(mci);
fail0:
iounmap(ovrfl_window);
pci_release_regions(ovrfl_pdev);
pci_disable_device(ovrfl_pdev);
/* NOTE: the ovrfl proc entry and pci_dev are intentionally left */
return rc;
}
/* returns count (>= 0), or negative on error */
static int i82875p_init_one(struct pci_dev *pdev,
const struct pci_device_id *ent)
{
int rc;
edac_dbg(0, "\n");
i82875p_printk(KERN_INFO, "i82875p init one\n");
if (pci_enable_device(pdev) < 0)
return -EIO;
rc = i82875p_probe1(pdev, ent->driver_data);
if (mci_pdev == NULL)
mci_pdev = pci_dev_get(pdev);
return rc;
}
static void i82875p_remove_one(struct pci_dev *pdev)
{
struct mem_ctl_info *mci;
struct i82875p_pvt *pvt = NULL;
edac_dbg(0, "\n");
if (i82875p_pci)
edac_pci_release_generic_ctl(i82875p_pci);
if ((mci = edac_mc_del_mc(&pdev->dev)) == NULL)
return;
pvt = (struct i82875p_pvt *)mci->pvt_info;
if (pvt->ovrfl_window)
iounmap(pvt->ovrfl_window);
if (pvt->ovrfl_pdev) {
#ifdef CORRECT_BIOS
pci_release_regions(pvt->ovrfl_pdev);
#endif /*CORRECT_BIOS */
pci_disable_device(pvt->ovrfl_pdev);
pci_dev_put(pvt->ovrfl_pdev);
}
edac_mc_free(mci);
}
static const struct pci_device_id i82875p_pci_tbl[] = {
{
PCI_VEND_DEV(INTEL, 82875_0), PCI_ANY_ID, PCI_ANY_ID, 0, 0,
I82875P},
{
0,
} /* 0 terminated list. */
};
MODULE_DEVICE_TABLE(pci, i82875p_pci_tbl);
static struct pci_driver i82875p_driver = {
.name = EDAC_MOD_STR,
.probe = i82875p_init_one,
.remove = i82875p_remove_one,
.id_table = i82875p_pci_tbl,
};
static int __init i82875p_init(void)
{
int pci_rc;
edac_dbg(3, "\n");
/* Ensure that the OPSTATE is set correctly for POLL or NMI */
opstate_init();
pci_rc = pci_register_driver(&i82875p_driver);
if (pci_rc < 0)
goto fail0;
if (mci_pdev == NULL) {
mci_pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
PCI_DEVICE_ID_INTEL_82875_0, NULL);
if (!mci_pdev) {
edac_dbg(0, "875p pci_get_device fail\n");
pci_rc = -ENODEV;
goto fail1;
}
pci_rc = i82875p_init_one(mci_pdev, i82875p_pci_tbl);
if (pci_rc < 0) {
edac_dbg(0, "875p init fail\n");
pci_rc = -ENODEV;
goto fail1;
}
}
return 0;
fail1:
pci_unregister_driver(&i82875p_driver);
fail0:
pci_dev_put(mci_pdev);
return pci_rc;
}
static void __exit i82875p_exit(void)
{
edac_dbg(3, "\n");
i82875p_edac: fix module remove Fix module removal bugs of i82875p_edac. Also i82975x_edac code seems to have the same module removal bugs as in i82875p_edac. The problems were: 1. In module removal i82875p_remove_one() is never called. Variable i82875p_registered is newer changed from 1, which guarantees i82875p_remove_one() is not called (and even if it were called, it would be called in wrong order). As a result, the edac_mc workque is not stopped and keeps probing. If kernel debugging options are not enabled, user may not notice anything going wrong. if debugging options are enabled and I do "rmmod i82875p_edac", I get: edac debug: edac_pci_workq_function() checking BUG: unable to handle kernel paging request at f882d16f ... call trace: [<f8834df3>] ? edac_mc_workq_function+0x55/0x7e [edac_core] [<c0233974>] ? run_workqueue+0xd7/0x1a5 [<c023392f>] ? run_workqueue+0x92/0x1a5 [<f8834d9e>] ? edac_mc_workq_function+0x0/0x7e [edac_core] [<c0233af9>] ? worker_thread+0xb7/0xc3 [<c0236a7b>] ? autoremove_wake_function+0x0/0x33 [<c0233a42>] ? worker_thread+0x0/0xc3 [<c0236809>] ? kthread+0x3b/0x61 [<c02367ce>] ? kthread+0x0/0x61 [<c0204587>] ? kernel_thread_helper+0x7/0x10 Fix for this is to get rid of needles variable i82875p_registered altogether and run i82875p_remove_one() *before* pci_unregister_driver(). 2. edac_mc_del_mc() uses mci after freeing mci edac_mc_del_mc() calls calls edac_remove_sysfs_mci_device(). The kobject refcount of mci drops to 0 and mci is freed. After this mci is accessed via debug print and i82875p_remove_one() still uses mci->pvt and tries to free mci again with edac_mc_free(). The fix for this is add kobject_get(&mci->edac_mci_kobj) after edac_mc_alloc(). Then the mci is still available after returning from edac_mc_del_mc() with refcount 1, and mci->pvt is still available. When i82875p_remove_one() finally calls edac_mc_free(), this will cause kobject_put() and mci is released properly. Signed-off-by: Jarkko Lavinen <jlavi@iki.fi> Cc: Doug Thompson <norsk5@yahoo.com> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-12-02 05:14:08 +08:00
i82875p_remove_one(mci_pdev);
pci_dev_put(mci_pdev);
pci_unregister_driver(&i82875p_driver);
}
module_init(i82875p_init);
module_exit(i82875p_exit);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Linux Networx (http://lnxi.com) Thayne Harbaugh");
MODULE_DESCRIPTION("MC support for Intel 82875 memory hub controllers");
module_param(edac_op_state, int, 0444);
MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");