From: Brad Peters <bpeters@redhat.com> Date: Fri, 8 Aug 2008 15:34:25 -0400 Subject: [ppc] edac: add support for Cell processor Message-id: 20080808193425.3203.53174.sendpatchset@squad5-lp1.lab.bos.redhat.com O-Subject: [PATCH 2/2] Bugzilla: 439507 RH-Acked-by: Aristeu Rozanski <arozansk@redhat.com> RHBZ#: ====== https://bugzilla.redhat.com/show_bug.cgi?id=439507 Description: =========== New Feature / Only affects Power (Cell, specifically) Patch 2/2 The Linux EDAC (Error Detection and Control) subsystem aims to tell you about problems with your hardware - problems which your hardware knows about, but which Linux has not previously checked for. This patch adds EDAC support to the Cell architecture. This patch actually adds the EDAC driver. Upstream Status: ================ commit 48764e4143c06672fc072eb482fdc4c75ee0f968 Author: Benjamin Herrenschmidt <benh@kernel.crashing.org> Date: Thu Feb 7 00:14:53 2008 -0800 drivers-edac: add Cell MC driver committer Benjamin Herrenschmidt <benh@kernel.crashing.org> commit c134fd868fb5137984ce1b7a58f3f6dd4171615e powerpc/cell/edac: Log a syndrome code in case of correctable error Test Status: ============ Patch set tested by confirming /sys/devices/system/edac* entries are present Brad Peters <IBM>, 8/8/08 =============================================================== Brad Peters 1-978-392-1000 x 23183 IBM on-site partner. Proposed Patch: =============== This patch is based on 2.6.18-102.el5 diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig index 31b2276..55a72f9 100644 --- a/drivers/edac/Kconfig +++ b/drivers/edac/Kconfig @@ -119,6 +119,14 @@ config EDAC_R82600 Support for error detection and correction on the Radisys 82600 embedded chipset. +config EDAC_CELL + tristate "Cell Broadband Engine memory controller" + depends on EDAC_MM_EDAC && PPC_CELL_NATIVE + help + Support for error detection and correction on the + Cell Broadband Engine internal memory controller + on platform without a hypervisor + choice prompt "Error detecting method" depends on EDAC diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile index fcb94ca..f47e393 100644 --- a/drivers/edac/Makefile +++ b/drivers/edac/Makefile @@ -18,4 +18,5 @@ obj-$(CONFIG_EDAC_I5000) += i5000_edac.o obj-$(CONFIG_EDAC_I82860) += i82860_edac.o obj-$(CONFIG_EDAC_K8) += k8_edac.o obj-$(CONFIG_EDAC_R82600) += r82600_edac.o +obj-$(CONFIG_EDAC_CELL) += cell_edac.o diff --git a/drivers/edac/cell_edac.c b/drivers/edac/cell_edac.c new file mode 100644 index 0000000..424387a --- /dev/null +++ b/drivers/edac/cell_edac.c @@ -0,0 +1,258 @@ +/* + * Cell MIC driver for ECC counting + * + * Copyright 2007 Benjamin Herrenschmidt, IBM Corp. + * <benh@kernel.crashing.org> + * + * This file may be distributed under the terms of the + * GNU General Public License. + */ +#undef DEBUG + +#include <linux/module.h> +#include <linux/init.h> +#include <linux/platform_device.h> +#include <linux/stop_machine.h> +#include <linux/io.h> +#include <asm/machdep.h> +#include <asm/cell-regs.h> + +#include "edac_mc.h" + +struct cell_edac_priv +{ + struct cbe_mic_tm_regs __iomem *regs; + int node; + int chanmask; +#ifdef DEBUG + u64 prev_fir; +#endif +}; + +static void cell_edac_count_ce(struct mem_ctl_info *mci, int chan, u64 ar) +{ + struct cell_edac_priv *priv = mci->pvt_info; + struct csrow_info *csrow = &mci->csrows[0]; + unsigned long address, pfn, offset, syndrome; + + dev_dbg(mci->dev, "ECC CE err on node %d, channel %d, ar = 0x%016lx\n", + priv->node, chan, ar); + + /* Address decoding is likely a bit bogus, to dbl check */ + address = (ar & 0xffffffffe0000000ul) >> 29; + if (priv->chanmask == 0x3) + address = (address << 1) | chan; + pfn = address >> PAGE_SHIFT; + offset = address & ~PAGE_MASK; + syndrome = (ar & 0x000000001fe00000ul) >> 21; + + /* TODO: Decoding of the error addresss */ + edac_mc_handle_ce(mci, csrow->first_page + pfn, offset, + syndrome, 0, chan, ""); +} + +static void cell_edac_count_ue(struct mem_ctl_info *mci, int chan, u64 ar) +{ + struct cell_edac_priv *priv = mci->pvt_info; + struct csrow_info *csrow = &mci->csrows[0]; + unsigned long address, pfn, offset; + + dev_dbg(mci->dev, "ECC UE err on node %d, channel %d, ar = 0x%016lx\n", + priv->node, chan, ar); + + /* Address decoding is likely a bit bogus, to dbl check */ + address = (ar & 0xffffffffe0000000ul) >> 29; + if (priv->chanmask == 0x3) + address = (address << 1) | chan; + pfn = address >> PAGE_SHIFT; + offset = address & ~PAGE_MASK; + + /* TODO: Decoding of the error addresss */ + edac_mc_handle_ue(mci, csrow->first_page + pfn, offset, 0, ""); +} + +static void cell_edac_check(struct mem_ctl_info *mci) +{ + struct cell_edac_priv *priv = mci->pvt_info; + u64 fir, addreg, clear = 0; + + fir = in_be64(&priv->regs->mic_fir); +#ifdef DEBUG + if (fir != priv->prev_fir) { + dev_dbg(mci->dev, "fir change : 0x%016lx\n", fir); + priv->prev_fir = fir; + } +#endif + if ((priv->chanmask & 0x1) && (fir & CBE_MIC_FIR_ECC_SINGLE_0_ERR)) { + addreg = in_be64(&priv->regs->mic_df_ecc_address_0); + clear |= CBE_MIC_FIR_ECC_SINGLE_0_RESET; + cell_edac_count_ce(mci, 0, addreg); + } + if ((priv->chanmask & 0x2) && (fir & CBE_MIC_FIR_ECC_SINGLE_1_ERR)) { + addreg = in_be64(&priv->regs->mic_df_ecc_address_1); + clear |= CBE_MIC_FIR_ECC_SINGLE_1_RESET; + cell_edac_count_ce(mci, 1, addreg); + } + if ((priv->chanmask & 0x1) && (fir & CBE_MIC_FIR_ECC_MULTI_0_ERR)) { + addreg = in_be64(&priv->regs->mic_df_ecc_address_0); + clear |= CBE_MIC_FIR_ECC_MULTI_0_RESET; + cell_edac_count_ue(mci, 0, addreg); + } + if ((priv->chanmask & 0x2) && (fir & CBE_MIC_FIR_ECC_MULTI_1_ERR)) { + addreg = in_be64(&priv->regs->mic_df_ecc_address_1); + clear |= CBE_MIC_FIR_ECC_MULTI_1_RESET; + cell_edac_count_ue(mci, 1, addreg); + } + + /* The procedure for clearing FIR bits is a bit ... weird */ + if (clear) { + fir &= ~(CBE_MIC_FIR_ECC_ERR_MASK | CBE_MIC_FIR_ECC_SET_MASK); + fir |= CBE_MIC_FIR_ECC_RESET_MASK; + fir &= ~clear; + out_be64(&priv->regs->mic_fir, fir); + (void)in_be64(&priv->regs->mic_fir); + + mb(); /* sync up */ +#ifdef DEBUG + fir = in_be64(&priv->regs->mic_fir); + dev_dbg(mci->dev, "fir clear : 0x%016lx\n", fir); +#endif + } +} + +static void __devinit cell_edac_init_csrows(struct mem_ctl_info *mci) +{ + struct csrow_info *csrow = &mci->csrows[0]; + struct cell_edac_priv *priv = mci->pvt_info; + struct device_node *np; + + for (np = NULL; + (np = of_find_node_by_name(np, "memory")) != NULL;) { + struct resource r; + + /* We "know" that the Cell firmware only creates one entry + * in the "memory" nodes. If that changes, this code will + * need to be adapted. + */ + if (of_address_to_resource(np, 0, &r)) + continue; + if (of_node_to_nid(np) != priv->node) + continue; + csrow->first_page = r.start >> PAGE_SHIFT; + csrow->nr_pages = (r.end - r.start + 1) >> PAGE_SHIFT; + csrow->last_page = csrow->first_page + csrow->nr_pages - 1; + csrow->mtype = MEM_XDR; + csrow->edac_mode = EDAC_FLAG_EC | EDAC_FLAG_SECDED; + dev_dbg(mci->dev, + "Initialized on node %d, chanmask=0x%x," + " first_page=0x%lx, nr_pages=0x%x\n", + priv->node, priv->chanmask, + csrow->first_page, csrow->nr_pages); + break; + } +} + +static int __devinit cell_edac_probe(struct platform_device *pdev) +{ + struct cbe_mic_tm_regs __iomem *regs; + struct mem_ctl_info *mci; + struct cell_edac_priv *priv; + u64 reg; + int rc, chanmask; + + regs = cbe_get_cpu_mic_tm_regs(cbe_node_to_cpu(pdev->id)); + if (regs == NULL) + return -ENODEV; + + /* Get channel population */ + reg = in_be64(®s->mic_mnt_cfg); + dev_dbg(&pdev->dev, "MIC_MNT_CFG = 0x%016lx\n", reg); + chanmask = 0; + if (reg & CBE_MIC_MNT_CFG_CHAN_0_POP) + chanmask |= 0x1; + if (reg & CBE_MIC_MNT_CFG_CHAN_1_POP) + chanmask |= 0x2; + if (chanmask == 0) { + dev_warn(&pdev->dev, + "Yuck ! No channel populated ? Aborting !\n"); + return -ENODEV; + } + dev_dbg(&pdev->dev, "Initial FIR = 0x%016lx\n", + in_be64(®s->mic_fir)); + + /* Allocate & init EDAC MC data structure */ + mci = edac_mc_alloc(sizeof(struct cell_edac_priv), 1, + chanmask == 3 ? 2 : 1); + if (mci == NULL) + return -ENOMEM; + priv = mci->pvt_info; + priv->regs = regs; + priv->node = pdev->id; + priv->chanmask = chanmask; + mci->dev = &pdev->dev; + mci->mtype_cap = MEM_FLAG_XDR; + mci->edac_ctl_cap = EDAC_FLAG_NONE | EDAC_FLAG_EC | EDAC_FLAG_SECDED; + mci->edac_cap = EDAC_FLAG_EC | EDAC_FLAG_SECDED; + mci->mod_name = "cell_edac"; + mci->ctl_name = "MIC"; + mci->edac_check = cell_edac_check; + cell_edac_init_csrows(mci); + + /* Register with EDAC core */ + rc = edac_mc_add_mc(mci, pdev->id); + if (rc) { + dev_err(&pdev->dev, "failed to register with EDAC core\n"); + edac_mc_free(mci); + return rc; + } + + return 0; +} + +static int __devexit cell_edac_remove(struct platform_device *pdev) +{ + struct mem_ctl_info *mci = edac_mc_del_mc(&pdev->dev); + if (mci) + edac_mc_free(mci); + return 0; +} + +static struct platform_driver cell_edac_driver = { + .driver = { + .name = "cbe-mic", + .owner = THIS_MODULE, + }, + .probe = cell_edac_probe, + .remove = cell_edac_remove, +}; + +static int __init cell_edac_init(void) +{ + /* Sanity check registers data structure */ + BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs, + mic_df_ecc_address_0) != 0xf8); + BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs, + mic_df_ecc_address_1) != 0x1b8); + BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs, + mic_df_config) != 0x218); + BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs, + mic_fir) != 0x230); + BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs, + mic_mnt_cfg) != 0x210); + BUILD_BUG_ON(offsetof(struct cbe_mic_tm_regs, + mic_exc) != 0x208); + + return platform_driver_register(&cell_edac_driver); +} + +static void __exit cell_edac_exit(void) +{ + platform_driver_unregister(&cell_edac_driver); +} + +module_init(cell_edac_init); +module_exit(cell_edac_exit); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Benjamin Herrenschmidt <benh@kernel.crashing.org>"); +MODULE_DESCRIPTION("ECC counting for Cell MIC");