Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S270867AbTG0XaA (ORCPT ); Sun, 27 Jul 2003 19:30:00 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S270539AbTG0XaA (ORCPT ); Sun, 27 Jul 2003 19:30:00 -0400 Received: from zeus.kernel.org ([204.152.189.113]:14071 "EHLO zeus.kernel.org") by vger.kernel.org with ESMTP id S273018AbTG0XDM (ORCPT ); Sun, 27 Jul 2003 19:03:12 -0400 From: Andrew de Quincey To: acpi-devel@lists.sourceforge.net, Andew Grover , lkml , torvalds@osdl.org Subject: [PATCH] ACPI patch which fixes all my IRQ problems on nforce2 Date: Sun, 27 Jul 2003 20:16:56 +0100 User-Agent: KMail/1.5.2 Cc: ollo.diab@gmx.de, Marcelo Penna Guerra MIME-Version: 1.0 Content-Type: Multipart/Mixed; boundary="Boundary-00=_oUCJ/+gD0XmJ84e" Message-Id: <200307272016.56573.adq_dvb@lidskialf.net> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 19514 Lines: 594 --Boundary-00=_oUCJ/+gD0XmJ84e Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Content-Disposition: inline Hi, this patch fixes all ACPI problems on the nforce2 chipset for me. Here is the feature list: 1) Make IO-APIC use correct polarity/mode settings from ACPI. 2) Detect buggy ACPI code when setting IRQ routing -- warn the user and attempt to fallback to non-ACPI configuration. 3) Add /proc/irq/io_apic so that the contents of the IO-APIC can be seen on a live system. 4) Fix potential problem in pci_link.c The current ACPI code does this: a) Scan for all PCI Link devices b) Allocate an IRQ to each link device. c) Setup PCI Routing table entries. This is error-prone. In my BIOS, there is one set of Link devices for PIC mode and one set for APIC mode. In the AML code however, they both talk to the *same* hardware on the board. So by luck, the PIC-mode link devices are set up first, then the APIC ones. If this order was reversed, the IRQs in APIC mode would be completely wrong. This patch changes the code so that it only allocates IRQs to link devices while setting up the PCI routing table entries, so only the PIC or APIC ones will be set up as appropriate. 5) Boot-enabled IRQs cannot be relied upon. Although the _CRS method in my ACPI does return the correct IRQ, this is NOT necessarily what the hardware is set to. The patch forces every link device IRQ to be set at least once. This _may_ break ACPI IRQ routing on the Toshiba 5005-S504 (I hope I have managed to support it though). Can someone check please? Quote from pci_link.c: * Note that we don't validate that the current IRQ (_CRS) exists * within the possible IRQs (_PRS): we blindly assume that whatever * IRQ a boot-enabled Link device is set to is the correct one. * (Required to support systems such as the Toshiba 5005-S504.) --Boundary-00=_oUCJ/+gD0XmJ84e Content-Type: text/x-diff; charset="us-ascii"; name="linux-2.5.75-acpi-irqparams-final3.patch" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="linux-2.5.75-acpi-irqparams-final3.patch" --- linux-2.5.75.orig/arch/i386/kernel/io_apic.c 2003-07-10 21:07:34.000000000 +0100 +++ linux-2.5.75/arch/i386/kernel/io_apic.c 2003-07-27 19:18:54.000000000 +0100 @@ -1544,6 +1544,142 @@ printk(KERN_DEBUG "... PIC ELCR: %04x\n", v); } +int io_apic_irq_read_proc (char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + int apic, i; + union IO_APIC_reg_00 reg_00; + union IO_APIC_reg_01 reg_01; + union IO_APIC_reg_02 reg_02; + union IO_APIC_reg_03 reg_03; + unsigned long flags; + char* start_page; + + start_page = page; + page += sprintf(page, "number of MP IRQ sources: %d.\n", mp_irq_entries); + for (i = 0; i < nr_ioapics; i++) + page += sprintf(page, "number of IO-APIC #%d registers: %d.\n", + mp_ioapics[i].mpc_apicid, nr_ioapic_registers[i]); + + for (apic = 0; apic < nr_ioapics; apic++) { + + spin_lock_irqsave(&ioapic_lock, flags); + reg_00.raw = io_apic_read(apic, 0); + reg_01.raw = io_apic_read(apic, 1); + if (reg_01.bits.version >= 0x10) + reg_02.raw = io_apic_read(apic, 2); + if (reg_01.bits.version >= 0x20) + reg_03.raw = io_apic_read(apic, 3); + spin_unlock_irqrestore(&ioapic_lock, flags); + + page += sprintf(page, "IO APIC #%d......\n", mp_ioapics[apic].mpc_apicid); + page += sprintf(page, ".... register #00: %08X\n", reg_00.raw); + page += sprintf(page, "....... : physical APIC id: %02X\n", reg_00.bits.ID); + page += sprintf(page, "....... : Delivery Type: %X\n", reg_00.bits.delivery_type); + page += sprintf(page, "....... : LTS : %X\n", reg_00.bits.LTS); + if (reg_00.bits.ID >= APIC_BROADCAST_ID) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + if (reg_00.bits.__reserved_1 || reg_00.bits.__reserved_2) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + + page += sprintf(page, ".... register #01: %08X\n", reg_01.raw); + page += sprintf(page, "....... : max redirection entries: %04X\n", reg_01.bits.entries); + if ( (reg_01.bits.entries != 0x0f) && /* older (Neptune) boards */ + (reg_01.bits.entries != 0x17) && /* typical ISA+PCI boards */ + (reg_01.bits.entries != 0x1b) && /* Compaq Proliant boards */ + (reg_01.bits.entries != 0x1f) && /* dual Xeon boards */ + (reg_01.bits.entries != 0x22) && /* bigger Xeon boards */ + (reg_01.bits.entries != 0x2E) && + (reg_01.bits.entries != 0x3F) + ) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + + page += sprintf(page, "....... : PRQ implemented: %X\n", reg_01.bits.PRQ); + page += sprintf(page, "....... : IO APIC version: %04X\n", reg_01.bits.version); + if ( (reg_01.bits.version != 0x01) && /* 82489DX IO-APICs */ + (reg_01.bits.version != 0x10) && /* oldest IO-APICs */ + (reg_01.bits.version != 0x11) && /* Pentium/Pro IO-APICs */ + (reg_01.bits.version != 0x13) && /* Xeon IO-APICs */ + (reg_01.bits.version != 0x20) /* Intel P64H (82806 AA) */ + ) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + if (reg_01.bits.__reserved_1 || reg_01.bits.__reserved_2) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + + /* + * Some Intel chipsets with IO APIC VERSION of 0x1? don't have reg_02, + * but the value of reg_02 is read as the previous read register + * value, so ignore it if reg_02 == reg_01. + */ + if (reg_01.bits.version >= 0x10 && reg_02.raw != reg_01.raw) { + page += sprintf(page, ".... register #02: %08X\n", reg_02.raw); + page += sprintf(page, "....... : arbitration: %02X\n", reg_02.bits.arbitration); + if (reg_02.bits.__reserved_1 || reg_02.bits.__reserved_2) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + } + + /* + * Some Intel chipsets with IO APIC VERSION of 0x2? don't have reg_02 + * or reg_03, but the value of reg_0[23] is read as the previous read + * register value, so ignore it if reg_03 == reg_0[12]. + */ + if (reg_01.bits.version >= 0x20 && reg_03.raw != reg_02.raw && + reg_03.raw != reg_01.raw) { + page += sprintf(page, ".... register #03: %08X\n", reg_03.raw); + page += sprintf(page, "....... : Boot DT : %X\n", reg_03.bits.boot_DT); + if (reg_03.bits.__reserved_1) + page += sprintf(page, "UNEXPECTED_IO_APIC\n"); + } + + page += sprintf(page, ".... IRQ redirection table:\n"); + + page += sprintf(page, " NR Log Phy Mask Trig IRR Pol" + " Stat Dest Deli Vect: \n"); + + for (i = 0; i <= reg_01.bits.entries; i++) { + struct IO_APIC_route_entry entry; + + spin_lock_irqsave(&ioapic_lock, flags); + *(((int *)&entry)+0) = io_apic_read(apic, 0x10+i*2); + *(((int *)&entry)+1) = io_apic_read(apic, 0x11+i*2); + spin_unlock_irqrestore(&ioapic_lock, flags); + + page += sprintf(page, " %02x %03X %02X ", + i, + entry.dest.logical.logical_dest, + entry.dest.physical.physical_dest + ); + + page += sprintf(page, "%1d %1d %1d %1d %1d %1d %1d %02X\n", + entry.mask, + entry.trigger, + entry.irr, + entry.polarity, + entry.delivery_status, + entry.dest_mode, + entry.delivery_mode, + entry.vector + ); + } + } + page += sprintf(page, "IRQ to pin mappings:\n"); + for (i = 0; i < NR_IRQS; i++) { + struct irq_pin_list *entry = irq_2_pin + i; + if (entry->pin < 0) + continue; + page += sprintf(page, "IRQ%d ", i); + for (;;) { + page += sprintf(page, "-> %d:0x%x", entry->apic, entry->pin); + if (!entry->next) + break; + entry = irq_2_pin + entry->next; + } + page += sprintf(page, "\n"); + } + + return page - start_page; +} + static void __init enable_IO_APIC(void) { union IO_APIC_reg_01 reg_01; @@ -2313,7 +2449,7 @@ } -int io_apic_set_pci_routing (int ioapic, int pin, int irq) +int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int active_high_low) { struct IO_APIC_route_entry entry; unsigned long flags; @@ -2335,19 +2471,23 @@ entry.delivery_mode = INT_DELIVERY_MODE; entry.dest_mode = INT_DEST_MODE; entry.dest.logical.logical_dest = cpu_mask_to_apicid(TARGET_CPUS); - entry.mask = 1; /* Disabled (masked) */ - entry.trigger = 1; /* Level sensitive */ - entry.polarity = 1; /* Low active */ + entry.trigger = edge_level; + entry.polarity = active_high_low; + entry.mask = 1; add_pin_to_irq(irq, ioapic, pin); entry.vector = assign_irq_vector(irq); printk(KERN_DEBUG "IOAPIC[%d]: Set PCI routing entry (%d-%d -> 0x%x -> " - "IRQ %d)\n", ioapic, - mp_ioapics[ioapic].mpc_apicid, pin, entry.vector, irq); + "IRQ %d Mode:%i Active:%i)\n", ioapic, + mp_ioapics[ioapic].mpc_apicid, pin, entry.vector, irq, edge_level, active_high_low); + if (edge_level) { irq_desc[irq].handler = &ioapic_level_irq_type; + } else { + irq_desc[irq].handler = &ioapic_edge_irq_type; + } set_intr_gate(entry.vector, interrupt[irq]); --- linux-2.5.75.orig/arch/i386/kernel/mpparse.c 2003-07-10 21:06:59.000000000 +0100 +++ linux-2.5.75/arch/i386/kernel/mpparse.c 2003-07-27 19:18:54.000000000 +0100 @@ -1065,7 +1065,7 @@ ioapic_pin = irq - mp_ioapic_routing[ioapic].irq_start; - io_apic_set_pci_routing(ioapic, ioapic_pin, irq); + io_apic_set_pci_routing(ioapic, ioapic_pin, irq, 1, 1); // Active low, edge triggered } #endif /*CONFIG_ACPI_HT_ONLY*/ @@ -1080,6 +1080,8 @@ int ioapic_pin = 0; int irq = 0; int idx, bit = 0; + int edge_level = 0; + int active_high_low = 0; /* * Parsing through the PCI Interrupt Routing Table (PRT) and program @@ -1090,12 +1092,16 @@ /* Need to get irq for dynamic entry */ if (entry->link.handle) { - irq = acpi_pci_link_get_irq(entry->link.handle, entry->link.index); + irq = acpi_pci_link_get_irq(entry->link.handle, entry->link.index, &edge_level, &active_high_low); if (!irq) continue; } - else + else { + /* Hardwired IRQ. Assume PCI standard settings */ irq = entry->link.index; + edge_level = 1; + active_high_low = 1; + } /* Don't set up the ACPI SCI because it's already set up */ if (acpi_fadt.sci_int == irq) @@ -1130,7 +1136,7 @@ mp_ioapic_routing[ioapic].pin_programmed[idx] |= (1<irq = irq; printk(KERN_DEBUG "%02x:%02x:%02x[%c] -> %d-%d -> IRQ %d\n", @@ -1140,6 +1146,9 @@ entry->irq); } + + + return; } --- linux-2.5.75.orig/arch/i386/kernel/irq.c 2003-07-10 21:04:50.000000000 +0100 +++ linux-2.5.75/arch/i386/kernel/irq.c 2003-07-27 19:18:54.000000000 +0100 @@ -1028,6 +1028,13 @@ unsigned long prof_cpu_mask = -1; +#ifdef CONFIG_X86_IO_APIC +static struct proc_dir_entry * io_apic_irq_entry; + +extern int io_apic_irq_read_proc (char *page, char **start, off_t off, + int count, int *eof, void *data); +#endif + void init_irq_proc (void) { struct proc_dir_entry *entry; @@ -1052,5 +1059,24 @@ */ for (i = 0; i < NR_IRQS; i++) register_irq_proc(i); + +#ifdef CONFIG_X86_IO_APIC + if (smp_found_config) { + if (!skip_ioapic_setup && nr_ioapics) { + struct proc_dir_entry* entry; + + /* create /proc/irq/io_apic */ + entry = create_proc_entry("io_apic", 0400, root_irq_dir); + + if (entry) { + entry->nlink = 1; + entry->read_proc = io_apic_irq_read_proc; + } + + io_apic_irq_entry = entry; + } + + } +#endif } --- linux-2.5.75.orig/drivers/acpi/pci_irq.c 2003-07-10 21:08:54.000000000 +0100 +++ linux-2.5.75/drivers/acpi/pci_irq.c 2003-07-27 19:18:54.000000000 +0100 @@ -249,7 +249,7 @@ } if (!entry->irq && entry->link.handle) { - entry->irq = acpi_pci_link_get_irq(entry->link.handle, entry->link.index); + entry->irq = acpi_pci_link_get_irq(entry->link.handle, entry->link.index, NULL, NULL); if (!entry->irq) { ACPI_DEBUG_PRINT((ACPI_DB_WARN, "Invalid IRQ link routing entry\n")); return_VALUE(0); @@ -389,7 +389,9 @@ } /* Make sure all link devices have a valid IRQ. */ - acpi_pci_link_check(); + if (acpi_pci_link_check()) { + return_VALUE(-ENODEV); + } #ifdef CONFIG_X86_IO_APIC /* Program IOAPICs using data from PRT entries. */ --- linux-2.5.75.orig/drivers/acpi/pci_link.c 2003-07-27 19:56:49.000000000 +0100 +++ linux-2.5.75/drivers/acpi/pci_link.c 2003-07-27 19:54:53.000000000 +0100 @@ -69,6 +69,9 @@ struct acpi_pci_link_irq { u8 active; /* Current IRQ */ + u8 edge_level; /* All IRQs */ + u8 active_high_low; /* All IRQs */ + u8 setonboot; u8 possible_count; u8 possible[ACPI_PCI_LINK_MAX_POSSIBLE]; }; @@ -118,6 +121,8 @@ link->irq.possible[i] = p->interrupts[i]; link->irq.possible_count++; } + link->irq.edge_level = p->edge_level; + link->irq.active_high_low = p->active_high_low; break; } case ACPI_RSTYPE_EXT_IRQ: @@ -136,6 +141,8 @@ link->irq.possible[i] = p->interrupts[i]; link->irq.possible_count++; } + link->irq.edge_level = p->edge_level; + link->irq.active_high_low = p->active_high_low; break; } default: @@ -264,7 +271,6 @@ * IRQ a boot-enabled Link device is set to is the correct one. * (Required to support systems such as the Toshiba 5005-S504.) */ - link->irq.active = irq; ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Link at IRQ %d \n", link->irq.active)); @@ -294,6 +300,8 @@ if (!link || !irq) return_VALUE(-EINVAL); + /* We don't check irqs the first time round */ + if (link->irq.setonboot) { /* See if we're already at the target IRQ. */ if (irq == link->irq.active) return_VALUE(0); @@ -307,16 +315,18 @@ ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Target IRQ %d invalid\n", irq)); return_VALUE(-EINVAL); } + } memset(&resource, 0, sizeof(resource)); - /* NOTE: PCI interrupts are always level / active_low / shared. */ + /* NOTE: PCI interrupts are always level / active_low / shared. But not all + interrupts > 15 are PCI interrupts. Rely on the ACPI IRQ definition for + parameters */ if (irq <= 15) { resource.res.id = ACPI_RSTYPE_IRQ; resource.res.length = sizeof(struct acpi_resource); - resource.res.data.irq.edge_level = ACPI_LEVEL_SENSITIVE; - resource.res.data.irq.active_high_low = ACPI_ACTIVE_LOW; - resource.res.data.irq.shared_exclusive = ACPI_SHARED; + resource.res.data.irq.edge_level = link->irq.edge_level; + resource.res.data.irq.active_high_low = link->irq.active_high_low; resource.res.data.irq.number_of_interrupts = 1; resource.res.data.irq.interrupts[0] = irq; } @@ -324,15 +334,15 @@ resource.res.id = ACPI_RSTYPE_EXT_IRQ; resource.res.length = sizeof(struct acpi_resource); resource.res.data.extended_irq.producer_consumer = ACPI_CONSUMER; - resource.res.data.extended_irq.edge_level = ACPI_LEVEL_SENSITIVE; - resource.res.data.extended_irq.active_high_low = ACPI_ACTIVE_LOW; - resource.res.data.extended_irq.shared_exclusive = ACPI_SHARED; + resource.res.data.extended_irq.edge_level = link->irq.edge_level; + resource.res.data.extended_irq.active_high_low = link->irq.active_high_low; resource.res.data.extended_irq.number_of_interrupts = 1; resource.res.data.extended_irq.interrupts[0] = irq; /* ignore resource_source, it's optional */ } resource.end.id = ACPI_RSTYPE_END_TAG; + /* Attempt to set the resource */ status = acpi_set_current_resources(link->handle, &buffer); if (ACPI_FAILURE(status)) { ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Error evaluating _SRS\n")); @@ -355,11 +365,13 @@ if (result) { return_VALUE(result); } + if (link->irq.active != irq) { ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Attempt to enable at IRQ %d resulted in IRQ %d\n", irq, link->irq.active)); link->irq.active = 0; + acpi_ut_evaluate_object (link->handle, "_DIS", 0, NULL); return_VALUE(-ENODEV); } @@ -407,7 +419,7 @@ ACPI_FUNCTION_TRACE("acpi_pci_link_check"); /* - * Pass #1: Update penalties to facilitate IRQ balancing. + * Update penalties to facilitate IRQ balancing. */ list_for_each(node, &acpi_link.entries) { @@ -428,22 +440,19 @@ } } - /* - * Pass #2: Enable boot-disabled Links at 'best' IRQ. - */ - list_for_each(node, &acpi_link.entries) { - int irq = 0; - int i = 0; + return_VALUE(0); +} - link = list_entry(node, struct acpi_pci_link, node); - if (!link || !link->irq.possible_count) { - ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Invalid link context\n")); - continue; - } +static int acpi_pci_link_allocate(struct acpi_pci_link* link) { + int irq; + int i; - if (link->irq.active) - continue; + if (link->irq.setonboot) + return_VALUE(0); + if (link->irq.active) { + irq = link->irq.active; + } else { irq = link->irq.possible[0]; /* @@ -454,17 +463,22 @@ if (acpi_irq_penalty[irq] > acpi_irq_penalty[link->irq.possible[i]]) irq = link->irq.possible[i]; } + } - /* Enable the link device at this IRQ. */ - acpi_pci_link_set(link, irq); - + /* Attempt to enable the link device at this IRQ. */ + if (acpi_pci_link_set(link, irq)) { + printk(PREFIX "Unable to set IRQ for %s [%s] (likely buggy ACPI BIOS). Aborting ACPI-based IRQ routing. Try pci=noacpi or acpi=off\n", + acpi_device_name(link->device), + acpi_device_bid(link->device)); + return_VALUE(-ENODEV); + } else { acpi_irq_penalty[link->irq.active] += 100; - printk(PREFIX "%s [%s] enabled at IRQ %d\n", acpi_device_name(link->device), acpi_device_bid(link->device), link->irq.active); } + link->irq.setonboot = 1; return_VALUE(0); } @@ -472,7 +486,9 @@ int acpi_pci_link_get_irq ( acpi_handle handle, - int index) + int index, + int* edge_level, + int* active_high_low) { int result = 0; struct acpi_device *device = NULL; @@ -498,11 +514,17 @@ return_VALUE(0); } + if (acpi_pci_link_allocate(link)) { + return -ENODEV; + } + if (!link->irq.active) { ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Link disabled\n")); return_VALUE(0); } + if (edge_level) *edge_level = link->irq.edge_level; + if (active_high_low) *active_high_low = link->irq.active_high_low; return_VALUE(link->irq.active); } --- linux-2.5.75.orig/include/acpi/acpi_drivers.h 2003-07-10 21:12:26.000000000 +0100 +++ linux-2.5.75/include/acpi/acpi_drivers.h 2003-07-27 19:18:54.000000000 +0100 @@ -60,7 +60,7 @@ /* ACPI PCI Interrupt Link (pci_link.c) */ int acpi_pci_link_check (void); -int acpi_pci_link_get_irq (acpi_handle handle, int index); +int acpi_pci_link_get_irq (acpi_handle handle, int index, int* edge_level, int* active_high_low); /* ACPI PCI Interrupt Routing (pci_irq.c) */ --- linux-2.5.75.orig/include/asm-i386/io_apic.h 2003-07-10 21:12:15.000000000 +0100 +++ linux-2.5.75/include/asm-i386/io_apic.h 2003-07-27 19:18:54.000000000 +0100 @@ -170,7 +170,7 @@ extern int io_apic_get_unique_id (int ioapic, int apic_id); extern int io_apic_get_version (int ioapic); extern int io_apic_get_redir_entries (int ioapic); -extern int io_apic_set_pci_routing (int ioapic, int pin, int irq); +extern int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int active_high_low); #endif /*CONFIG_ACPI_BOOT*/ #else /* !CONFIG_X86_IO_APIC */ --Boundary-00=_oUCJ/+gD0XmJ84e-- - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/