From patchwork Mon Nov 15 07:30:42 2010 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Isaku Yamahata X-Patchwork-Id: 71197 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from lists.gnu.org (lists.gnu.org [199.232.76.165]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTPS id 05DF9B7108 for ; Mon, 15 Nov 2010 18:45:31 +1100 (EST) Received: from localhost ([127.0.0.1]:32919 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1PHtkQ-0002rP-5w for incoming@patchwork.ozlabs.org; Mon, 15 Nov 2010 02:45:06 -0500 Received: from [140.186.70.92] (port=51247 helo=eggs.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1PHtWk-0003fX-Jj for qemu-devel@nongnu.org; Mon, 15 Nov 2010 02:31:06 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1PHtWf-00027N-A1 for qemu-devel@nongnu.org; Mon, 15 Nov 2010 02:30:58 -0500 Received: from mail.valinux.co.jp ([210.128.90.3]:48006) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1PHtWe-000268-2W for qemu-devel@nongnu.org; Mon, 15 Nov 2010 02:30:53 -0500 Received: from ps.local.valinux.co.jp (vagw.valinux.co.jp [210.128.90.14]) by mail.valinux.co.jp (Postfix) with SMTP id 49F4B8746B; Mon, 15 Nov 2010 16:30:47 +0900 (JST) Received: (nullmailer pid 10371 invoked by uid 1000); Mon, 15 Nov 2010 07:30:47 -0000 From: Isaku Yamahata To: qemu-devel@nongnu.org Date: Mon, 15 Nov 2010 16:30:42 +0900 Message-Id: <00e4193aa6cb5ab56ca554f248713c497f1e3af2.1289805831.git.yamahata@valinux.co.jp> X-Mailer: git-send-email 1.7.1.1 In-Reply-To: References: In-Reply-To: References: X-Virus-Scanned: clamav-milter 0.95.2 at va-mail.local.valinux.co.jp X-Virus-Status: Clean X-detected-operating-system: by eggs.gnu.org: GNU/Linux 2.6 (newer, 3) Cc: skandasa@cisco.com, adnan@khaleel.us, wexu2@cisco.com, mst@redhat.com, yamahata@valinux.co.jp, etmartin@cisco.com Subject: [Qemu-devel] [PATCH v8 06/11] pcie/aer: helper functions for pcie aer capability X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org This patch implements helper functions for pcie aer capability which will be used later. Signed-off-by: Isaku Yamahata --- Changes v7 -> v8: - various changes to follow the review. - s/TLP_PRESENT/TLP_PREFIX_PRESENT/g Changes v6 -> v7: - make error log buffer non-ringed buffer - refactor pcie_aer_error_inject() - clean up of pcie_aer_write_config() - make pcie_aer_inject_error() return error - remove AERMsgResult - remove PCIEAERSeverity - reduce forward declarations - style clean ups - remove paren for sizeof. Chnages v5 -> v6: - cleaned up pcie_aer_write_config(). - enum definition. Changes v4 -> v5: - use pci_xxx_test_and_xxx_mask() - rewrote PCIDevice::written bits. - eliminated pcie_aer_notify() - introduced PCIExpressDevice::aer_intx Changes v3 -> v4: - various naming fixes. - use pci bit operation helper function - eliminate errmsg function pointer - replace pci_shift_xxx() with PCIDevice::written - uncorrect error status register. - dropped pcie_aer_cap() Changes v2 -> v3: - split out from pcie.[ch] to pcie_aer.[ch] to make the files sorter. - embeded PCIExpressDevice into PCIDevice. - CodingStyle fix --- Makefile.objs | 2 +- hw/pcie.h | 14 + hw/pcie_aer.c | 826 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++ hw/pcie_aer.h | 106 ++++++++ qemu-common.h | 3 + 5 files changed, 950 insertions(+), 1 deletions(-) create mode 100644 hw/pcie_aer.c create mode 100644 hw/pcie_aer.h diff --git a/Makefile.objs b/Makefile.objs index 15569af..6f01220 100644 --- a/Makefile.objs +++ b/Makefile.objs @@ -208,7 +208,7 @@ hw-obj-$(CONFIG_PIIX4) += piix4.o # PCI watchdog devices hw-obj-y += wdt_i6300esb.o -hw-obj-y += pcie.o pcie_port.o +hw-obj-y += pcie.o pcie_aer.o pcie_port.o hw-obj-y += msix.o msi.o # PCI network cards diff --git a/hw/pcie.h b/hw/pcie.h index 8708504..7baa813 100644 --- a/hw/pcie.h +++ b/hw/pcie.h @@ -24,6 +24,7 @@ #include "hw.h" #include "pci_regs.h" #include "pcie_regs.h" +#include "pcie_aer.h" typedef enum { /* for attention and power indicator */ @@ -79,6 +80,19 @@ struct PCIExpressDevice { Software Notification of Hot-Plug Events, an interrupt is sent whenever the logical and of these conditions transitions from false to true. */ + + /* AER */ + uint16_t aer_cap; + PCIEAERLog aer_log; + unsigned int aer_intx; /* INTx for error reporting + * default is 0 = INTA# + * If the chip wants to use other interrupt + * line, initialize this member with the + * desired number. + * If the chip dynamically changes this member, + * also initialize it when loaded as + * appropreately. + */ }; /* PCI express capability helper functions */ diff --git a/hw/pcie_aer.c b/hw/pcie_aer.c new file mode 100644 index 0000000..0aabdb3 --- /dev/null +++ b/hw/pcie_aer.c @@ -0,0 +1,826 @@ +/* + * pcie_aer.c + * + * Copyright (c) 2010 Isaku Yamahata + * VA Linux Systems Japan K.K. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, see . + */ + +#include "sysemu.h" +#include "pci_bridge.h" +#include "pcie.h" +#include "msix.h" +#include "msi.h" +#include "pci_internals.h" +#include "pcie_regs.h" + +//#define DEBUG_PCIE +#ifdef DEBUG_PCIE +# define PCIE_DPRINTF(fmt, ...) \ + fprintf(stderr, "%s:%d " fmt, __func__, __LINE__, ## __VA_ARGS__) +#else +# define PCIE_DPRINTF(fmt, ...) do {} while (0) +#endif +#define PCIE_DEV_PRINTF(dev, fmt, ...) \ + PCIE_DPRINTF("%s:%x "fmt, (dev)->name, (dev)->devfn, ## __VA_ARGS__) + +/* From 6.2.7 Error Listing and Rules. Table 6-2, 6-3 and 6-4 */ +static uint32_t pcie_aer_uncor_default_severity(uint32_t status) +{ + switch (status) { + case PCI_ERR_UNC_INTN: + case PCI_ERR_UNC_DLP: + case PCI_ERR_UNC_SDN: + case PCI_ERR_UNC_RX_OVER: + case PCI_ERR_UNC_FCP: + case PCI_ERR_UNC_MALF_TLP: + return PCI_ERR_ROOT_CMD_FATAL_EN; + case PCI_ERR_UNC_POISON_TLP: + case PCI_ERR_UNC_ECRC: + case PCI_ERR_UNC_UNSUP: + case PCI_ERR_UNC_COMP_TIME: + case PCI_ERR_UNC_COMP_ABORT: + case PCI_ERR_UNC_UNX_COMP: + case PCI_ERR_UNC_ACSV: + case PCI_ERR_UNC_MCBTLP: + case PCI_ERR_UNC_ATOP_EBLOCKED: + case PCI_ERR_UNC_TLP_PRF_BLOCKED: + return PCI_ERR_ROOT_CMD_NONFATAL_EN; + default: + abort(); + break; + } + return PCI_ERR_ROOT_CMD_FATAL_EN; +} + +static int aer_log_add_err(PCIEAERLog *aer_log, const PCIEAERErr *err) +{ + if (aer_log->log_num == aer_log->log_max) { + return -1; + } + memcpy(&aer_log->log[aer_log->log_num], err, sizeof *err); + aer_log->log_num++; + return 0; +} + +static void aer_log_del_err(PCIEAERLog *aer_log, PCIEAERErr *err) +{ + assert(aer_log->log_num); + *err = aer_log->log[0]; + aer_log->log_num--; + memmove(&aer_log->log[0], &aer_log->log[1], + aer_log->log_num * sizeof *err); +} + +static void aer_log_clear_all_err(PCIEAERLog *aer_log) +{ + aer_log->log_num = 0; +} + +void pcie_aer_init(PCIDevice *dev, uint16_t offset) +{ + PCIExpressDevice *exp; + + pcie_add_capability(dev, PCI_EXT_CAP_ID_ERR, PCI_ERR_VER, + offset, PCI_ERR_SIZEOF); + exp = &dev->exp; + exp->aer_cap = offset; + + /* log_max is property */ + if (dev->exp.aer_log.log_max == PCIE_AER_LOG_MAX_UNSET) { + dev->exp.aer_log.log_max = PCIE_AER_LOG_MAX_DEFAULT; + } + /* clip down the value to avoid unreasobale memory usage */ + if (dev->exp.aer_log.log_max > PCIE_AER_LOG_MAX_LIMIT) { + dev->exp.aer_log.log_max = PCIE_AER_LOG_MAX_LIMIT; + } + dev->exp.aer_log.log = qemu_mallocz(sizeof dev->exp.aer_log.log[0] * + dev->exp.aer_log.log_max); + + pci_set_long(dev->w1cmask + offset + PCI_ERR_UNCOR_STATUS, + PCI_ERR_UNC_SUPPORTED); + + pci_set_long(dev->config + offset + PCI_ERR_UNCOR_SEVER, + PCI_ERR_UNC_SEVERITY_DEFAULT); + pci_set_long(dev->wmask + offset + PCI_ERR_UNCOR_SEVER, + PCI_ERR_UNC_SUPPORTED); + + pci_long_test_and_set_mask(dev->w1cmask + offset + PCI_ERR_COR_STATUS, + PCI_ERR_COR_STATUS); + + pci_set_long(dev->config + offset + PCI_ERR_COR_MASK, + PCI_ERR_COR_MASK_DEFAULT); + pci_set_long(dev->wmask + offset + PCI_ERR_COR_MASK, + PCI_ERR_COR_SUPPORTED); + + /* capabilities and control. multiple header logging is supported */ + if (dev->exp.aer_log.log_max > 0) { + pci_set_long(dev->config + offset + PCI_ERR_CAP, + PCI_ERR_CAP_ECRC_GENC | PCI_ERR_CAP_ECRC_CHKC | + PCI_ERR_CAP_MHRC); + pci_set_long(dev->wmask + offset + PCI_ERR_CAP, + PCI_ERR_CAP_ECRC_GENE | PCI_ERR_CAP_ECRC_CHKE | + PCI_ERR_CAP_MHRE); + } else { + pci_set_long(dev->config + offset + PCI_ERR_CAP, + PCI_ERR_CAP_ECRC_GENC | PCI_ERR_CAP_ECRC_CHKC); + pci_set_long(dev->wmask + offset + PCI_ERR_CAP, + PCI_ERR_CAP_ECRC_GENE | PCI_ERR_CAP_ECRC_CHKE); + } + + switch (pcie_cap_get_type(dev)) { + case PCI_EXP_TYPE_ROOT_PORT: + /* this case will be set by pcie_aer_root_init() */ + /* fallthrough */ + case PCI_EXP_TYPE_DOWNSTREAM: + case PCI_EXP_TYPE_UPSTREAM: + pci_word_test_and_set_mask(dev->wmask + PCI_BRIDGE_CONTROL, + PCI_BRIDGE_CTL_SERR); + pci_long_test_and_set_mask(dev->w1cmask + PCI_STATUS, + PCI_SEC_STATUS_RCV_SYSTEM_ERROR); + break; + default: + /* nothing */ + break; + } +} + +void pcie_aer_exit(PCIDevice *dev) +{ + qemu_free(dev->exp.aer_log.log); +} + +static void pcie_aer_update_uncor_status(PCIDevice *dev) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + PCIEAERLog *aer_log = &dev->exp.aer_log; + + uint16_t i; + for (i = 0; i < aer_log->log_num; i++) { + pci_long_test_and_set_mask(aer_cap + PCI_ERR_UNCOR_STATUS, + dev->exp.aer_log.log[i].status); + } +} + +/* + * pcie_aer_msg() is called recursively by + * pcie_aer_msg_alldev(), pci_aer_msg_vbridge() and pcie_aer_msg_root_port() + */ +static void pcie_aer_msg(PCIDevice *dev, const PCIEAERMsg *msg); + +/* + * return value: + * true: error message is sent up + * false: error message is masked + * + * 6.2.6 Error Message Control + * Figure 6-3 + * all pci express devices part + */ +static bool +pcie_aer_msg_alldev(PCIDevice *dev, const PCIEAERMsg *msg) +{ + PCIDevice *parent_port; + + if (!(pcie_aer_msg_is_uncor(msg) && + (pci_get_word(dev->config + PCI_COMMAND) & PCI_COMMAND_SERR))) { + return false; + } + + /* Signaled System Error + * + * 7.5.1.1 Command register + * Bit 8 SERR# Enable + * + * When Set, this bit enables reporting of Non-fatal and Fatal + * errors detected by the Function to the Root Complex. Note that + * errors are reported if enabled either through this bit or through + * the PCI Express specific bits in the Device Control register (see + * Section 7.8.4). + */ + pci_word_test_and_set_mask(dev->config + PCI_STATUS, + PCI_STATUS_SIG_SYSTEM_ERROR); + + if (!(msg->severity & + pci_get_word(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL))) { + return false; + } + + /* send up error message */ + if (pci_is_express(dev) && + pcie_cap_get_type(dev) == PCI_EXP_TYPE_ROOT_PORT) { + /* Root port notify system itself, + or send the error message to root complex event collector. */ + /* + * if root port is associated to event collector, set + * parent_port = root complex event collector + * For now root complex event collector isn't supported. + */ + parent_port = NULL; + } else { + parent_port = pci_bridge_get_device(dev->bus); + } + if (parent_port) { + if (!pci_is_express(parent_port)) { + /* just ignore it */ + return false; + } + pcie_aer_msg(parent_port, msg); + } + return true; +} + +/* + * return value: + * true: error message is sent up + * false: error message is masked + * + * 6.2.6 Error Message Control + * Figure 6-3 + * virtual pci bridge part + */ +static bool pcie_aer_msg_vbridge(PCIDevice *dev, const PCIEAERMsg *msg) +{ + uint16_t bridge_control = pci_get_word(dev->config + PCI_BRIDGE_CONTROL); + + if (pcie_aer_msg_is_uncor(msg)) { + /* Received System Error */ + pci_word_test_and_set_mask(dev->config + PCI_SEC_STATUS, + PCI_SEC_STATUS_RCV_SYSTEM_ERROR); + } + + if (!(bridge_control & PCI_BRIDGE_CTL_SERR)) { + return false; + } + return true; +} + +void pcie_aer_root_set_vector(PCIDevice *dev, unsigned int vector) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + assert(vector < PCI_ERR_ROOT_IRQ_MAX); + pci_long_test_and_clear_mask(aer_cap + PCI_ERR_ROOT_STATUS, + PCI_ERR_ROOT_IRQ); + pci_long_test_and_set_mask(aer_cap + PCI_ERR_ROOT_STATUS, + vector << PCI_ERR_ROOT_IRQ_SHIFT); +} + +static unsigned int pcie_aer_root_get_vector(PCIDevice *dev) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + uint32_t root_status = pci_get_long(aer_cap + PCI_ERR_ROOT_STATUS); + return (root_status & PCI_ERR_ROOT_IRQ) >> PCI_ERR_ROOT_IRQ_SHIFT; +} + +/* + * return value: + * true: error message is sent up + * false: error message is masked + * + * 6.2.6 Error Message Control + * Figure 6-3 + * root port part + */ +static bool pcie_aer_msg_root_port(PCIDevice *dev, const PCIEAERMsg *msg) +{ + bool msg_sent; + uint16_t cmd; + uint8_t *aer_cap; + uint32_t root_cmd; + uint32_t root_status; + bool msi_trigger; + + msg_sent = false; + cmd = pci_get_word(dev->config + PCI_COMMAND); + aer_cap = dev->config + dev->exp.aer_cap; + root_cmd = pci_get_long(aer_cap + PCI_ERR_ROOT_COMMAND); + root_status = pci_get_long(aer_cap + PCI_ERR_ROOT_STATUS); + msi_trigger = false; + + if (cmd & PCI_COMMAND_SERR) { + /* System Error. + * + * The way to report System Error is platform specific and + * it isn't implemented in qemu right now. + * So just discard the error for now. + * OS which cares of aer would receive errors via + * native aer mechanims, so this wouldn't matter. + */ + } + + /* Errro Message Received: Root Error Status register */ + switch (msg->severity) { + case PCI_ERR_ROOT_CMD_COR_EN: + if (root_status & PCI_ERR_ROOT_COR_RCV) { + root_status |= PCI_ERR_ROOT_MULTI_COR_RCV; + } else { + if (root_cmd & PCI_ERR_ROOT_CMD_COR_EN) { + msi_trigger = true; + } + pci_set_word(aer_cap + PCI_ERR_ROOT_COR_SRC, msg->source_id); + } + root_status |= PCI_ERR_ROOT_COR_RCV; + break; + case PCI_ERR_ROOT_CMD_NONFATAL_EN: + if (!(root_status & PCI_ERR_ROOT_NONFATAL_RCV) && + root_cmd & PCI_ERR_ROOT_CMD_NONFATAL_EN) { + msi_trigger = true; + } + root_status |= PCI_ERR_ROOT_NONFATAL_RCV; + break; + case PCI_ERR_ROOT_CMD_FATAL_EN: + if (!(root_status & PCI_ERR_ROOT_FATAL_RCV) && + root_cmd & PCI_ERR_ROOT_CMD_FATAL_EN) { + msi_trigger = true; + } + if (!(root_status & PCI_ERR_ROOT_UNCOR_RCV)) { + root_status |= PCI_ERR_ROOT_FIRST_FATAL; + } + root_status |= PCI_ERR_ROOT_FATAL_RCV; + break; + default: + abort(); + break; + } + if (pcie_aer_msg_is_uncor(msg)) { + if (root_status & PCI_ERR_ROOT_UNCOR_RCV) { + root_status |= PCI_ERR_ROOT_MULTI_UNCOR_RCV; + } else { + pci_set_word(aer_cap + PCI_ERR_ROOT_SRC, msg->source_id); + } + root_status |= PCI_ERR_ROOT_UNCOR_RCV; + } + pci_set_long(aer_cap + PCI_ERR_ROOT_STATUS, root_status); + + if (root_cmd & msg->severity) { + /* 6.2.4.1.2 Interrupt Generation */ + if (pci_msi_enabled(dev)) { + if (msi_trigger) { + pci_msi_notify(dev, pcie_aer_root_get_vector(dev)); + } + } else { + qemu_set_irq(dev->irq[dev->exp.aer_intx], 1); + } + msg_sent = true; + } + return msg_sent; +} + +/* + * 6.2.6 Error Message Control Figure 6-3 + */ +static void pcie_aer_msg(PCIDevice *dev, const PCIEAERMsg *msg) +{ + uint8_t type; + bool msg_sent; + + assert(pci_is_express(dev)); + + type = pcie_cap_get_type(dev); + if (type == PCI_EXP_TYPE_ROOT_PORT || + type == PCI_EXP_TYPE_UPSTREAM || + type == PCI_EXP_TYPE_DOWNSTREAM) { + msg_sent = pcie_aer_msg_vbridge(dev, msg); + if (!msg_sent) { + return; + } + } + msg_sent = pcie_aer_msg_alldev(dev, msg); + if (type == PCI_EXP_TYPE_ROOT_PORT && msg_sent) { + pcie_aer_msg_root_port(dev, msg); + } +} + +static void pcie_aer_update_log(PCIDevice *dev, const PCIEAERErr *err) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + uint8_t first_bit = ffsl(err->status) - 1; + uint32_t errcap = pci_get_long(aer_cap + PCI_ERR_CAP); + int i; + + assert(err->status); + assert(err->status & (err->status - 1)); + + errcap &= ~(PCI_ERR_CAP_FEP_MASK | PCI_ERR_CAP_TLP); + errcap |= PCI_ERR_CAP_FEP(first_bit); + + if (err->flags & PCIE_AER_ERR_HEADER_VALID) { + for (i = 0; i < ARRAY_SIZE(err->header); ++i) { + /* 7.10.8 Header Log Register */ + uint8_t *header_log = + aer_cap + PCI_ERR_HEADER_LOG + i * sizeof err->header[0]; + cpu_to_be32wu((uint32_t*)header_log, err->header[i]); + } + } else { + assert(!(err->flags & PCIE_AER_ERR_TLP_PREFIX_PRESENT)); + memset(aer_cap + PCI_ERR_HEADER_LOG, 0, PCI_ERR_HEADER_LOG_SIZE); + } + + if ((err->flags & PCIE_AER_ERR_TLP_PREFIX_PRESENT) && + (pci_get_long(dev->config + dev->exp.exp_cap + PCI_EXP_DEVCTL2) & + PCI_EXP_DEVCAP2_EETLPP)) { + for (i = 0; i < ARRAY_SIZE(err->prefix); ++i) { + /* 7.10.12 tlp prefix log register */ + uint8_t *prefix_log = + aer_cap + PCI_ERR_TLP_PREFIX_LOG + i * sizeof err->prefix[0]; + cpu_to_be32wu((uint32_t*)prefix_log, err->prefix[i]); + } + errcap |= PCI_ERR_CAP_TLP; + } else { + memset(aer_cap + PCI_ERR_TLP_PREFIX_LOG, 0, + PCI_ERR_TLP_PREFIX_LOG_SIZE); + } + pci_set_long(aer_cap + PCI_ERR_CAP, errcap); +} + +static void pcie_aer_clear_log(PCIDevice *dev) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + + pci_long_test_and_clear_mask(aer_cap + PCI_ERR_CAP, + PCI_ERR_CAP_FEP_MASK | PCI_ERR_CAP_TLP); + + memset(aer_cap + PCI_ERR_HEADER_LOG, 0, PCI_ERR_HEADER_LOG_SIZE); + memset(aer_cap + PCI_ERR_TLP_PREFIX_LOG, 0, PCI_ERR_TLP_PREFIX_LOG_SIZE); +} + +static void pcie_aer_clear_error(PCIDevice *dev) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + uint32_t errcap = pci_get_long(aer_cap + PCI_ERR_CAP); + PCIEAERLog *aer_log = &dev->exp.aer_log; + PCIEAERErr err; + + if (!(errcap & PCI_ERR_CAP_MHRE) || !aer_log->log_num) { + pcie_aer_clear_log(dev); + return; + } + + /* + * If more errors are queued, set corresponding bits in uncorrectable + * error status. + * We emulate uncorrectable error status register as W1CS. + * So set bit in uncorrectable error status here again for multiple + * error recording support. + * + * 6.2.4.2 Multiple Error Handling(Advanced Error Reporting Capability) + */ + pcie_aer_update_uncor_status(dev); + + aer_log_del_err(aer_log, &err); + pcie_aer_update_log(dev, &err); +} + +static int pcie_aer_record_error(PCIDevice *dev, + const PCIEAERErr *err) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + uint32_t errcap = pci_get_long(aer_cap + PCI_ERR_CAP); + int fep = PCI_ERR_CAP_FEP(errcap); + + assert(err->status); + assert(err->status & (err->status - 1)); + + if (errcap & PCI_ERR_CAP_MHRE && + (pci_get_long(aer_cap + PCI_ERR_UNCOR_STATUS) & (1U << fep))) { + /* Not first error. queue error */ + if (aer_log_add_err(&dev->exp.aer_log, err) < 0) { + /* overflow */ + return -1; + } + return 0; + } + + pcie_aer_update_log(dev, err); + return 0; +} + +typedef struct PCIEAERInject { + PCIDevice *dev; + uint8_t *aer_cap; + const PCIEAERErr *err; + uint16_t devctl; + uint16_t devsta; + uint32_t error_status; + bool unsupported_request; + bool log_overflow; + PCIEAERMsg msg; +} PCIEAERInject; + +static bool pcie_aer_inject_cor_error(PCIEAERInject *inj, + uint32_t uncor_status, + bool is_advisory_nonfatal) +{ + PCIDevice *dev = inj->dev; + + inj->devsta |= PCI_EXP_DEVSTA_CED; + if (inj->unsupported_request) { + inj->devsta |= PCI_EXP_DEVSTA_URD; + } + pci_set_word(dev->config + dev->exp.exp_cap + PCI_EXP_DEVSTA, inj->devsta); + + if (inj->aer_cap) { + uint32_t mask; + pci_long_test_and_set_mask(inj->aer_cap + PCI_ERR_COR_STATUS, + inj->error_status); + mask = pci_get_long(inj->aer_cap + PCI_ERR_COR_MASK); + if (mask & inj->error_status) { + return false; + } + if (is_advisory_nonfatal) { + uint32_t uncor_mask = + pci_get_long(inj->aer_cap + PCI_ERR_UNCOR_MASK); + if (!(uncor_mask & uncor_status)) { + inj->log_overflow = !!pcie_aer_record_error(dev, inj->err); + } + pci_long_test_and_set_mask(inj->aer_cap + PCI_ERR_UNCOR_STATUS, + uncor_status); + } + } + + if (inj->unsupported_request && !(inj->devctl & PCI_EXP_DEVCTL_URRE)) { + return false; + } + if (!(inj->devctl & PCI_EXP_DEVCTL_CERE)) { + return false; + } + + inj->msg.severity = PCI_ERR_ROOT_CMD_COR_EN; + return true; +} + +static bool pcie_aer_inject_uncor_error(PCIEAERInject *inj, bool is_fatal) +{ + PCIDevice *dev = inj->dev; + uint16_t cmd; + + if (is_fatal) { + inj->devsta |= PCI_EXP_DEVSTA_FED; + } else { + inj->devsta |= PCI_EXP_DEVSTA_NFED; + } + if (inj->unsupported_request) { + inj->devsta |= PCI_EXP_DEVSTA_URD; + } + pci_set_long(dev->config + dev->exp.exp_cap + PCI_EXP_DEVSTA, inj->devsta); + + if (inj->aer_cap) { + uint32_t mask = pci_get_long(inj->aer_cap + PCI_ERR_UNCOR_MASK); + if (mask & inj->error_status) { + pci_long_test_and_set_mask(inj->aer_cap + PCI_ERR_UNCOR_STATUS, + inj->error_status); + return false; + } + + inj->log_overflow = !!pcie_aer_record_error(dev, inj->err); + pci_long_test_and_set_mask(inj->aer_cap + PCI_ERR_UNCOR_STATUS, + inj->error_status); + } + + cmd = pci_get_word(dev->config + PCI_COMMAND); + if (inj->unsupported_request && + !(inj->devctl & PCI_EXP_DEVCTL_URRE) && !(cmd & PCI_COMMAND_SERR)) { + return false; + } + if (is_fatal) { + if (!((cmd & PCI_COMMAND_SERR) || + (inj->devctl & PCI_EXP_DEVCTL_FERE))) { + return false; + } + inj->msg.severity = PCI_ERR_ROOT_CMD_FATAL_EN; + } else { + if (!((cmd & PCI_COMMAND_SERR) || + (inj->devctl & PCI_EXP_DEVCTL_NFERE))) { + return false; + } + inj->msg.severity = PCI_ERR_ROOT_CMD_NONFATAL_EN; + } + return true; +} + +/* + * non-Function specific error must be recorded in all functions. + * It is the responsibility of the caller of this function. + * It is also caller's responsiblity to determine which function should + * report the rerror. + * + * 6.2.4 Error Logging + * 6.2.5 Sqeunce of Device Error Signaling and Logging Operations + * table 6-2: Flowchard Showing Sequence of Device Error Signaling and Logging + * Operations + */ +int pcie_aer_inject_error(PCIDevice *dev, const PCIEAERErr *err) +{ + uint8_t *aer_cap = NULL; + uint16_t devctl = 0; + uint16_t devsta = 0; + uint32_t error_status = err->status; + PCIEAERInject inj; + + if (!pci_is_express(dev)) { + return -1; + } + + if (err->flags & PCIE_AER_ERR_IS_CORRECTABLE) { + error_status &= PCI_ERR_COR_SUPPORTED; + } else { + error_status &= PCI_ERR_UNC_SUPPORTED; + } + + /* invalid status bit. one and only one bit must be set */ + if (!error_status || (error_status & (error_status - 1))) { + return -1; + } + + if (dev->exp.aer_cap) { + uint8_t *exp_cap = dev->config + dev->exp.exp_cap; + aer_cap = dev->config + dev->exp.aer_cap; + devctl = pci_get_long(exp_cap + PCI_EXP_DEVCTL); + devsta = pci_get_long(exp_cap + PCI_EXP_DEVSTA); + } + + inj.dev = dev; + inj.aer_cap = aer_cap; + inj.err = err; + inj.devctl = devctl; + inj.devsta = devsta; + inj.error_status = error_status; + inj.unsupported_request = !(err->flags & PCIE_AER_ERR_IS_CORRECTABLE) && + err->status == PCI_ERR_UNC_UNSUP; + inj.log_overflow = false; + + if (err->flags & PCIE_AER_ERR_IS_CORRECTABLE) { + if (!pcie_aer_inject_cor_error(&inj, 0, false)) { + return 0; + } + } else { + bool is_fatal = + pcie_aer_uncor_default_severity(error_status) == + PCI_ERR_ROOT_CMD_FATAL_EN; + if (aer_cap) { + is_fatal = + error_status & pci_get_long(aer_cap + PCI_ERR_UNCOR_SEVER); + } + if (!is_fatal && (err->flags & PCIE_AER_ERR_MAYBE_ADVISORY)) { + inj.error_status = PCI_ERR_COR_ADV_NONFATAL; + if (!pcie_aer_inject_cor_error(&inj, error_status, true)) { + return 0; + } + } else { + if (!pcie_aer_inject_uncor_error(&inj, is_fatal)) { + return 0; + } + } + } + + /* send up error message */ + inj.msg.source_id = err->source_id; + pcie_aer_msg(dev, &inj.msg); + + if (inj.log_overflow) { + PCIEAERErr header_log_overflow = { + .status = PCI_ERR_COR_HL_OVERFLOW, + .flags = PCIE_AER_ERR_IS_CORRECTABLE, + }; + int ret = pcie_aer_inject_error(dev, &header_log_overflow); + assert(!ret); + } + return 0; +} + +void pcie_aer_write_config(PCIDevice *dev, + uint32_t addr, uint32_t val, int len) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + uint32_t errcap = pci_get_long(aer_cap + PCI_ERR_CAP); + uint32_t first_error = 1U << PCI_ERR_CAP_FEP(errcap); + uint32_t uncorsta = pci_get_long(aer_cap + PCI_ERR_UNCOR_STATUS); + + /* uncorrectable error */ + if (!(uncorsta & first_error)) { + /* the bit that corresponds to the first error is cleared */ + pcie_aer_clear_error(dev); + } else if (errcap & PCI_ERR_CAP_MHRE) { + /* When PCI_ERR_CAP_MHRE is enabled and the first error isn't cleared + * nothing should happen. So we have to revert the modification to + * the register. + */ + pcie_aer_update_uncor_status(dev); + } else { + /* capability & control + * PCI_ERR_CAP_MHRE might be cleared, so clear of header log. + */ + aer_log_clear_all_err(&dev->exp.aer_log); + } +} + +void pcie_aer_root_init(PCIDevice *dev) +{ + uint16_t pos = dev->exp.aer_cap; + + pci_set_long(dev->wmask + pos + PCI_ERR_ROOT_COMMAND, + PCI_ERR_ROOT_CMD_EN_MASK); + pci_set_long(dev->w1cmask + pos + PCI_ERR_ROOT_STATUS, + PCI_ERR_ROOT_STATUS_REPORT_MASK); +} + +void pcie_aer_root_reset(PCIDevice *dev) +{ + uint8_t* aer_cap = dev->config + dev->exp.aer_cap; + + pci_set_long(aer_cap + PCI_ERR_ROOT_COMMAND, 0); + + /* + * Advanced Error Interrupt Message Number in Root Error Status Register + * must be updated by chip dependent code because it's chip dependent + * which number is used. + */ +} + +static bool pcie_aer_root_does_trigger(uint32_t cmd, uint32_t status) +{ + return + ((cmd & PCI_ERR_ROOT_CMD_COR_EN) && (status & PCI_ERR_ROOT_COR_RCV)) || + ((cmd & PCI_ERR_ROOT_CMD_NONFATAL_EN) && + (status & PCI_ERR_ROOT_NONFATAL_RCV)) || + ((cmd & PCI_ERR_ROOT_CMD_FATAL_EN) && + (status & PCI_ERR_ROOT_FATAL_RCV)); +} + +void pcie_aer_root_write_config(PCIDevice *dev, + uint32_t addr, uint32_t val, int len, + uint32_t root_cmd_prev) +{ + uint8_t *aer_cap = dev->config + dev->exp.aer_cap; + + /* root command register */ + uint32_t root_cmd = pci_get_long(aer_cap + PCI_ERR_ROOT_COMMAND); + if (root_cmd & PCI_ERR_ROOT_CMD_EN_MASK) { + /* 6.2.4.1.2 Interrupt Generation */ + + /* 0 -> 1 */ + uint32_t root_cmd_set = (root_cmd_prev ^ root_cmd) & root_cmd; + uint32_t root_status = pci_get_long(aer_cap + PCI_ERR_ROOT_STATUS); + + if (pci_msi_enabled(dev)) { + if (pcie_aer_root_does_trigger(root_cmd_set, root_status)) { + pci_msi_notify(dev, pcie_aer_root_get_vector(dev)); + } + } else { + int int_level = pcie_aer_root_does_trigger(root_cmd, root_status); + qemu_set_irq(dev->irq[dev->exp.aer_intx], int_level); + } + } +} + +static const VMStateDescription vmstate_pcie_aer_err = { + .name = "PCIE_AER_ERROR", + .version_id = 1, + .minimum_version_id = 1, + .minimum_version_id_old = 1, + .fields = (VMStateField[]) { + VMSTATE_UINT32(status, PCIEAERErr), + VMSTATE_UINT16(source_id, PCIEAERErr), + VMSTATE_UINT16(flags, PCIEAERErr), + VMSTATE_UINT32_ARRAY(header, PCIEAERErr, 4), + VMSTATE_UINT32_ARRAY(prefix, PCIEAERErr, 4), + VMSTATE_END_OF_LIST() + } +}; + +#define VMSTATE_PCIE_AER_ERRS(_field, _state, _field_num, _vmsd, _type) { \ + .name = (stringify(_field)), \ + .version_id = 0, \ + .num_offset = vmstate_offset_value(_state, _field_num, uint16_t), \ + .size = sizeof(_type), \ + .vmsd = &(_vmsd), \ + .flags = VMS_POINTER | VMS_VARRAY_UINT16 | VMS_STRUCT, \ + .offset = vmstate_offset_pointer(_state, _field, _type), \ +} + +const VMStateDescription vmstate_pcie_aer_log = { + .name = "PCIE_AER_ERROR_LOG", + .version_id = 1, + .minimum_version_id = 1, + .minimum_version_id_old = 1, + .fields = (VMStateField[]) { + VMSTATE_UINT16(log_num, PCIEAERLog), + VMSTATE_UINT16(log_max, PCIEAERLog), + VMSTATE_PCIE_AER_ERRS(log, PCIEAERLog, log_num, + vmstate_pcie_aer_err, PCIEAERErr), + VMSTATE_END_OF_LIST() + } +}; + diff --git a/hw/pcie_aer.h b/hw/pcie_aer.h new file mode 100644 index 0000000..3de7763 --- /dev/null +++ b/hw/pcie_aer.h @@ -0,0 +1,106 @@ +/* + * pcie_aer.h + * + * Copyright (c) 2010 Isaku Yamahata + * VA Linux Systems Japan K.K. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, see . + */ + +#ifndef QEMU_PCIE_AER_H +#define QEMU_PCIE_AER_H + +#include "hw.h" + +/* definitions which PCIExpressDevice uses */ + +/* AER log */ +struct PCIEAERLog { + /* This structure is saved/loaded. + So explicitly size them instead of unsigned int */ + + /* the number of currently recorded log in log member */ + uint16_t log_num; + + /* + * The maximum number of the log. Errors can be logged up to this. + * + * This is configurable property. + * The specified value will be clipped down to PCIE_AER_LOG_MAX_LIMIT + * to avoid unreasonable memory usage. + * I bet that 128 log size would be big enough, otherwise too many errors + * for system to function normaly. But could consecutive errors occur? + */ +#define PCIE_AER_LOG_MAX_DEFAULT 8 +#define PCIE_AER_LOG_MAX_LIMIT 128 +#define PCIE_AER_LOG_MAX_UNSET 0xffff + uint16_t log_max; + + /* Error log. log_max-sized array */ + PCIEAERErr *log; +}; + +/* aer error message: error signaling message has only error sevirity and + source id. See 2.2.8.3 error signaling messages */ +struct PCIEAERMsg { + /* + * PCI_ERR_ROOT_CMD_{COR, NONFATAL, FATAL}_EN + * = PCI_EXP_DEVCTL_{CERE, NFERE, FERE} + */ + uint32_t severity; + + uint16_t source_id; /* bdf */ +}; + +static inline bool +pcie_aer_msg_is_uncor(const PCIEAERMsg *msg) +{ + return msg->severity == PCI_ERR_ROOT_CMD_NONFATAL_EN || + msg->severity == PCI_ERR_ROOT_CMD_FATAL_EN; +} + +/* error */ +struct PCIEAERErr { + uint32_t status; /* error status bits */ + uint16_t source_id; /* bdf */ + +#define PCIE_AER_ERR_IS_CORRECTABLE 0x1 /* correctable/uncorrectable */ +#define PCIE_AER_ERR_MAYBE_ADVISORY 0x2 /* maybe advisory non-fatal */ +#define PCIE_AER_ERR_HEADER_VALID 0x4 /* TLP header is logged */ +#define PCIE_AER_ERR_TLP_PREFIX_PRESENT 0x8 /* TLP Prefix is logged */ + uint16_t flags; + + uint32_t header[4]; /* TLP header */ + uint32_t prefix[4]; /* TLP header prefix */ +}; + +extern const VMStateDescription vmstate_pcie_aer_log; + +void pcie_aer_init(PCIDevice *dev, uint16_t offset); +void pcie_aer_exit(PCIDevice *dev); +void pcie_aer_write_config(PCIDevice *dev, + uint32_t addr, uint32_t val, int len); + +/* aer root port */ +void pcie_aer_root_set_vector(PCIDevice *dev, unsigned int vector); +void pcie_aer_root_init(PCIDevice *dev); +void pcie_aer_root_reset(PCIDevice *dev); +void pcie_aer_root_write_config(PCIDevice *dev, + uint32_t addr, uint32_t val, int len, + uint32_t root_cmd_prev); + +/* error injection */ +int pcie_aer_inject_error(PCIDevice *dev, const PCIEAERErr *err); + +#endif /* QEMU_PCIE_AER_H */ diff --git a/qemu-common.h b/qemu-common.h index b3957f1..de82c2e 100644 --- a/qemu-common.h +++ b/qemu-common.h @@ -240,6 +240,9 @@ typedef struct PCIBus PCIBus; typedef struct PCIDevice PCIDevice; typedef struct PCIExpressDevice PCIExpressDevice; typedef struct PCIBridge PCIBridge; +typedef struct PCIEAERMsg PCIEAERMsg; +typedef struct PCIEAERLog PCIEAERLog; +typedef struct PCIEAERErr PCIEAERErr; typedef struct PCIEPort PCIEPort; typedef struct PCIESlot PCIESlot; typedef struct SerialState SerialState;