From owner-svn-src-all@freebsd.org Wed Jun 19 06:41:08 2019 Return-Path: Delivered-To: svn-src-all@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 0764A15D3CF5; Wed, 19 Jun 2019 06:41:08 +0000 (UTC) (envelope-from scottl@FreeBSD.org) Received: from mxrelay.nyi.freebsd.org (mxrelay.nyi.freebsd.org [IPv6:2610:1c1:1:606c::19:3]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) server-signature RSA-PSS (4096 bits) client-signature RSA-PSS (4096 bits) client-digest SHA256) (Client CN "mxrelay.nyi.freebsd.org", Issuer "Let's Encrypt Authority X3" (verified OK)) by mx1.freebsd.org (Postfix) with ESMTPS id A1C7777831; Wed, 19 Jun 2019 06:41:07 +0000 (UTC) (envelope-from scottl@FreeBSD.org) Received: from repo.freebsd.org (repo.freebsd.org [IPv6:2610:1c1:1:6068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mxrelay.nyi.freebsd.org (Postfix) with ESMTPS id 7E99D20AF1; Wed, 19 Jun 2019 06:41:07 +0000 (UTC) (envelope-from scottl@FreeBSD.org) Received: from repo.freebsd.org ([127.0.1.37]) by repo.freebsd.org (8.15.2/8.15.2) with ESMTP id x5J6f7OV006328; Wed, 19 Jun 2019 06:41:07 GMT (envelope-from scottl@FreeBSD.org) Received: (from scottl@localhost) by repo.freebsd.org (8.15.2/8.15.2/Submit) id x5J6f7ej006327; Wed, 19 Jun 2019 06:41:07 GMT (envelope-from scottl@FreeBSD.org) Message-Id: <201906190641.x5J6f7ej006327@repo.freebsd.org> X-Authentication-Warning: repo.freebsd.org: scottl set sender to scottl@FreeBSD.org using -f From: Scott Long Date: Wed, 19 Jun 2019 06:41:07 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r349184 - head/sys/amd64/vmm/intel X-SVN-Group: head X-SVN-Commit-Author: scottl X-SVN-Commit-Paths: head/sys/amd64/vmm/intel X-SVN-Commit-Revision: 349184 X-SVN-Commit-Repository: base MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-Rspamd-Queue-Id: A1C7777831 X-Spamd-Bar: -- Authentication-Results: mx1.freebsd.org X-Spamd-Result: default: False [-2.97 / 15.00]; local_wl_from(0.00)[FreeBSD.org]; NEURAL_HAM_MEDIUM(-1.00)[-0.999,0]; NEURAL_HAM_SHORT(-0.97)[-0.966,0]; ASN(0.00)[asn:11403, ipnet:2610:1c1:1::/48, country:US]; NEURAL_HAM_LONG(-1.00)[-1.000,0] X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 19 Jun 2019 06:41:08 -0000 Author: scottl Date: Wed Jun 19 06:41:07 2019 New Revision: 349184 URL: https://svnweb.freebsd.org/changeset/base/349184 Log: Implement VT-d capability detection on chipsets that have multiple translation units with differing capabilities From the author via Bugzilla: --- When an attempt is made to passthrough a PCI device to a bhyve VM (causing initialisation of IOMMU) on certain Intel chipsets using VT-d the PCI bus stops working entirely. This issue occurs on the E3-1275 v5 processor on C236 chipset and has also been encountered by others on the forums with different hardware in the Skylake series. The chipset has two VT-d translation units. The issue is caused by an attempt to use the VT-d device-IOTLB capability that is supported by only the first unit for devices attached to the second unit which lacks that capability. Only the capabilities of the first unit are checked and are assumed to be the same for all units. Attached is a patch to rectify this issue by determining which unit is responsible for the device being added to a domain and then checking that unit's device-IOTLB capability. In addition to this a few fixes have been made to other instances where the first unit's capabilities are assumed for all units for domains they share. In these cases a mutual set of capabilities is determined. The patch should hopefully fix any bugs for current/future hardware with multiple translation units supporting different capabilities. A description is on the forums at https://forums.freebsd.org/threads/pci-passthrough-bhyve-usb-xhci.65235 The thread includes observations by other users of the bug occurring, and description as well as confirmation of the fix. I'd also like to thank Ordoban for their help. --- Personally tested on a Skylake laptop, Skylake Xeon server, and a Xeon-D-1541, passing through XHCI and NVMe functions. Passthru is hit-or-miss to the point of being unusable without this patch. PR: 229852 Submitted by: callum@aitchison.org MFC after: 1 week Modified: head/sys/amd64/vmm/intel/vtd.c Modified: head/sys/amd64/vmm/intel/vtd.c ============================================================================== --- head/sys/amd64/vmm/intel/vtd.c Wed Jun 19 03:33:00 2019 (r349183) +++ head/sys/amd64/vmm/intel/vtd.c Wed Jun 19 06:41:07 2019 (r349184) @@ -51,6 +51,8 @@ __FBSDID("$FreeBSD$"); * Architecture Spec, September 2008. */ +#define VTD_DRHD_INCLUDE_PCI_ALL(Flags) (((Flags) >> 0) & 0x1) + /* Section 10.4 "Register Descriptions" */ struct vtdmap { volatile uint32_t version; @@ -116,10 +118,11 @@ struct domain { static SLIST_HEAD(, domain) domhead; #define DRHD_MAX_UNITS 8 -static int drhd_num; -static struct vtdmap *vtdmaps[DRHD_MAX_UNITS]; -static int max_domains; -typedef int (*drhd_ident_func_t)(void); +static ACPI_DMAR_HARDWARE_UNIT *drhds[DRHD_MAX_UNITS]; +static int drhd_num; +static struct vtdmap *vtdmaps[DRHD_MAX_UNITS]; +static int max_domains; +typedef int (*drhd_ident_func_t)(void); static uint64_t root_table[PAGE_SIZE / sizeof(uint64_t)] __aligned(4096); static uint64_t ctx_tables[256][PAGE_SIZE / sizeof(uint64_t)] __aligned(4096); @@ -175,6 +178,69 @@ domain_id(void) return (id); } +static struct vtdmap * +vtd_device_scope(uint16_t rid) +{ + int i, remaining, pathremaining; + char *end, *pathend; + struct vtdmap *vtdmap; + ACPI_DMAR_HARDWARE_UNIT *drhd; + ACPI_DMAR_DEVICE_SCOPE *device_scope; + ACPI_DMAR_PCI_PATH *path; + + for (i = 0; i < drhd_num; i++) { + drhd = drhds[i]; + + if (VTD_DRHD_INCLUDE_PCI_ALL(drhd->Flags)) { + /* + * From Intel VT-d arch spec, version 3.0: + * If a DRHD structure with INCLUDE_PCI_ALL flag Set is reported + * for a Segment, it must be enumerated by BIOS after all other + * DRHD structures for the same Segment. + */ + vtdmap = vtdmaps[i]; + return(vtdmap); + } + + end = (char *)drhd + drhd->Header.Length; + remaining = drhd->Header.Length - sizeof(ACPI_DMAR_HARDWARE_UNIT); + while (remaining > sizeof(ACPI_DMAR_DEVICE_SCOPE)) { + device_scope = (ACPI_DMAR_DEVICE_SCOPE *)(end - remaining); + remaining -= device_scope->Length; + + switch (device_scope->EntryType){ + /* 0x01 and 0x02 are PCI device entries */ + case 0x01: + case 0x02: + break; + default: + continue; + } + + if (PCI_RID2BUS(rid) != device_scope->Bus) + continue; + + pathend = (char *)device_scope + device_scope->Length; + pathremaining = device_scope->Length - sizeof(ACPI_DMAR_DEVICE_SCOPE); + while (pathremaining >= sizeof(ACPI_DMAR_PCI_PATH)) { + path = (ACPI_DMAR_PCI_PATH *)(pathend - pathremaining); + pathremaining -= sizeof(ACPI_DMAR_PCI_PATH); + + if (PCI_RID2SLOT(rid) != path->Device) + continue; + if (PCI_RID2FUNC(rid) != path->Function) + continue; + + vtdmap = vtdmaps[i]; + return (vtdmap); + } + } + } + + /* No matching scope */ + return (NULL); +} + static void vtd_wbflush(struct vtdmap *vtdmap) { @@ -240,7 +306,7 @@ vtd_translation_disable(struct vtdmap *vtdmap) static int vtd_init(void) { - int i, units, remaining; + int i, units, remaining, tmp; struct vtdmap *vtdmap; vm_paddr_t ctx_paddr; char *end, envname[32]; @@ -291,8 +357,9 @@ vtd_init(void) break; drhd = (ACPI_DMAR_HARDWARE_UNIT *)hdr; - vtdmaps[units++] = (struct vtdmap *)PHYS_TO_DMAP(drhd->Address); - if (units >= DRHD_MAX_UNITS) + drhds[units] = drhd; + vtdmaps[units] = (struct vtdmap *)PHYS_TO_DMAP(drhd->Address); + if (++units >= DRHD_MAX_UNITS) break; remaining -= hdr->Length; } @@ -302,13 +369,19 @@ vtd_init(void) skip_dmar: drhd_num = units; - vtdmap = vtdmaps[0]; - if (VTD_CAP_CM(vtdmap->cap) != 0) - panic("vtd_init: invalid caching mode"); + max_domains = 64 * 1024; /* maximum valid value */ + for (i = 0; i < drhd_num; i++){ + vtdmap = vtdmaps[i]; - max_domains = vtd_max_domains(vtdmap); + if (VTD_CAP_CM(vtdmap->cap) != 0) + panic("vtd_init: invalid caching mode"); + /* take most compatible (minimum) value */ + if ((tmp = vtd_max_domains(vtdmap)) < max_domains) + max_domains = tmp; + } + /* * Set up the root-table to point to the context-entry tables */ @@ -373,7 +446,6 @@ vtd_add_device(void *arg, uint16_t rid) struct vtdmap *vtdmap; uint8_t bus; - vtdmap = vtdmaps[0]; bus = PCI_RID2BUS(rid); ctxp = ctx_tables[bus]; pt_paddr = vtophys(dom->ptp); @@ -385,6 +457,10 @@ vtd_add_device(void *arg, uint16_t rid) (uint16_t)(ctxp[idx + 1] >> 8)); } + if ((vtdmap = vtd_device_scope(rid)) == NULL) + panic("vtd_add_device: device %x is not in scope for " + "any DMA remapping unit", rid); + /* * Order is important. The 'present' bit is set only after all fields * of the context pointer are initialized. @@ -568,8 +644,6 @@ vtd_create_domain(vm_paddr_t maxaddr) if (drhd_num <= 0) panic("vtd_create_domain: no dma remapping hardware available"); - vtdmap = vtdmaps[0]; - /* * Calculate AGAW. * Section 3.4.2 "Adjusted Guest Address Width", Architecture Spec. @@ -594,7 +668,14 @@ vtd_create_domain(vm_paddr_t maxaddr) pt_levels = 2; sagaw = 30; addrwidth = 0; - tmp = VTD_CAP_SAGAW(vtdmap->cap); + + tmp = ~0; + for (i = 0; i < drhd_num; i++) { + vtdmap = vtdmaps[i]; + /* take most compatible value */ + tmp &= VTD_CAP_SAGAW(vtdmap->cap); + } + for (i = 0; i < 5; i++) { if ((tmp & (1 << i)) != 0 && sagaw >= agaw) break; @@ -606,8 +687,8 @@ vtd_create_domain(vm_paddr_t maxaddr) } if (i >= 5) { - panic("vtd_create_domain: SAGAW 0x%lx does not support AGAW %d", - VTD_CAP_SAGAW(vtdmap->cap), agaw); + panic("vtd_create_domain: SAGAW 0x%x does not support AGAW %d", + tmp, agaw); } dom = malloc(sizeof(struct domain), M_VTD, M_ZERO | M_WAITOK); @@ -634,7 +715,12 @@ vtd_create_domain(vm_paddr_t maxaddr) * There is not any code to deal with the demotion at the moment * so we disable superpage mappings altogether. */ - dom->spsmask = VTD_CAP_SPS(vtdmap->cap); + dom->spsmask = ~0; + for (i = 0; i < drhd_num; i++) { + vtdmap = vtdmaps[i]; + /* take most compatible value */ + dom->spsmask &= VTD_CAP_SPS(vtdmap->cap); + } #endif SLIST_INSERT_HEAD(&domhead, dom, next);