On 5/21/20 11:01 AM, Laine Stump wrote:
On 5/8/20 4:54 PM, Jim Fehlig wrote:On 4/24/20 9:07 PM, Marek Marczykowski-Górecki wrote:From: Simon Gaiser <simon@xxxxxxxxxxxxxxxxxxxxxx> By setting the permissive flag the Xen guest access to the PCI config space is not filtered. This might be a security risk, but it's required for some devices and the IOMMU and interrupt remapping should (mostly?) contain it. Because of it (and that the attribute is Xen only), in an example include "permissive='no'" - to not expose users copying from documentation to extra risk. Example usage: <devices> ... <hostdev mode='subsystem' type='pci' managed='yes' permissive='yes'>As you know I always struggle with XML modeling and naming, and IIRC the 'managed' attribute has been a source of pain in the past, so it would be nice to have someone else opine on the introduction of 'permissive' attribute. I see Laine reviewed V1 and as I recall has commented on the managed attribute in the past, so let's see if he has any advice :-).Sorry, this message got lost in the firehose, and Jim just now reminded me of it on IRC.
Hi Laine, Thanks for you comments!
I seriously don't remember what I may have said in the past about the managed='yes' attribute. Could have been something about the default being 'yes' instead of 'no', or could have been about its placement as an attribute of the toplevel <hostdev> element rather than being made a part of the <source> subelement (since it is really associated with what is done on the host side to get the device setup and hooked into the qemu process).For this "permissive" attribute, I don't really know enough about it to know where it should go. Does it affect how the device appears in the guest? (in which case maybe it should be a part of <target>) or only the trappings around it in the host? (so maybe it should be a part of <source>). If not totally one or the other, could the setting of this option be changed and guest ABI would remain unchanged (but just some operations would now fail / succeed when they had done the opposite in the past)? In that case, again maybe <source>.
I don't follow xen development enough these days to confidently answer your questions. I'm quite sure the setting doesn't affect how the device appears in the guest, only some filtering of PCI config spaces access. Marek, can you verify that? If so, I lean towards your suggestion of adding the attribute to <source>.
I guess the answers to that would help guide me if I was making the decision about where to put the new setting.As for the *name* of it? That's even more difficult for me to have any opinion on since, again, I really have no understanding of what it's doing or why it's necessary.
From the xl.cfg(5) man page:By default pciback only allows PV guests to write "known safe" values into PCI configuration space, likewise QEMU (both qemu-xen and qemu-xen-traditional) imposes the same constraint on HVM guests. However, many devices require writes to other areas of the configuration space in order to operate properly. This option tells the backend (pciback or QEMU) to allow all writes to the PCI configuration space of this device by this domain.
It sounds more and more like a config option for the <source>. Do any of these options sound plausible?
<source permissive='on|off'>
<address .../>
</source>
<source write-filtering='on|off'>
<address .../>
</source>
<source>
<address .../>
<config-space filtering='on|off'/>
</source>
Regards,
Jim
One thing I would ask (maybe I did before) - wherever you put it, we (people who use qemu) would *really* appreciate if you added a clause to the post-parse validation in the qemu driver to cause an error if this option is found in a qemu domain config :-)
I think you mentioned it in V1. Marek, can you add such a check in the qemu driver for V3?
Regards, Jim
One option I considered was 'filtered' and 'unfiltered' values for a "config space" attribute, similar to sgio for scsi hostdevs. But I can't think of a good name for an attribute that "filters writes to the PCI hostdev config space".<source> <address domain='0x0000' bus='0x06' slot='0x02' function='0x0'/> </source> </hostdev> </devices> Signed-off-by: Simon Gaiser <simon@xxxxxxxxxxxxxxxxxxxxxx> Signed-off-by: Marek Marczykowski-Górecki <marmarek@xxxxxxxxxxxxxxxxxxxxxx> --- Changes in v2: - improve documentation (version info, example) - update schema - use virTristateBool - add a test - improve commit message - news entry --- docs/formatdomain.html.in | 7 +++++-- docs/news.xml | 11 +++++++++++-Additions to news is usually done in a sparate patch.docs/schemas/domaincommon.rng | 10 ++++++++++- src/conf/domain_conf.c | 19 +++++++++++++++++++- src/conf/domain_conf.h | 1 +- src/libxl/libxl_conf.c | 1 +- tests/libxlxml2domconfigdata/moredevs-hvm.json | 6 ++++++- tests/libxlxml2domconfigdata/moredevs-hvm.xml | 5 +++++- 8 files changed, 58 insertions(+), 2 deletions(-) diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index c530573..0d1146e 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -4987,7 +4987,7 @@ <pre> ... <devices> - <hostdev mode='subsystem' type='pci' managed='yes'> + <hostdev mode='subsystem' type='pci' managed='yes' permissive='no'> <source> <address domain='0x0000' bus='0x06' slot='0x02' function='0x0'/> </source> @@ -5082,7 +5082,10 @@ (or <code>virsh nodedev-detach</code> before starting the guest or hot-plugging the device and <code>virNodeDeviceReAttach</code> (or <code>virsh nodedev-reattach</code>) after hot-unplug or - stopping the guest. + stopping the guest. When <code>permissive</code> + (<span class="since">since 6.3.0, Xen only</span>) is "yes"6.4.0 Regards, Jim+ the pci config space access will not be filtered. This might be + a security issue. The default is "no". </dd> <dt><code>scsi</code></dt> <dd>For SCSI devices, user is responsible to make sure the device diff --git a/docs/news.xml b/docs/news.xml index 5835013..a8e992a 100644 --- a/docs/news.xml +++ b/docs/news.xml @@ -109,6 +109,17 @@ and/or fine tuned per individual host. </description> </change> + <change> + <summary> + xen: Add support for 'permissive' PCI device option + </summary> + <description> + <code>permissive</code> is a Xen-specific PCI device option that + disables config space write filtering. It is needed for proper + functioning of some devices using non-standard config space + registers. + </description> + </change> </section> <section title="Improvements"> </section> diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng index 7f18e5b..5a71222 100644 --- a/docs/schemas/domaincommon.rng +++ b/docs/schemas/domaincommon.rng @@ -3065,6 +3065,11 @@ <ref name="virYesNo"/> </attribute> </optional> + <optional> + <attribute name="permissive"> + <ref name="virYesNo"/> + </attribute> + </optional> <interleave> <element name="source"> <optional> @@ -4899,6 +4904,11 @@ <ref name="virYesNo"/> </attribute> </optional> + <optional> + <attribute name="permissive"> + <ref name="virYesNo"/> + </attribute> + </optional> <choice> <ref name="hostdevsubsyspci"/> <ref name="hostdevsubsysusb"/> diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 89cd8c5..fe1a864 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -8434,6 +8434,7 @@ virDomainHostdevDefParseXMLSubsys(xmlNodePtr node,virDomainHostdevSubsysSCSIVHostPtr scsihostsrc = &def->source.subsys.u.scsi_host;virDomainHostdevSubsysMediatedDevPtr mdevsrc = &def->source.subsys.u.mdev; g_autofree char *managed = NULL; + g_autofree char *permissive = NULL; g_autofree char *sgio = NULL; g_autofree char *rawio = NULL; g_autofree char *backendStr = NULL; @@ -8449,6 +8450,15 @@ virDomainHostdevDefParseXMLSubsys(xmlNodePtr node, if ((managed = virXMLPropString(node, "managed")) != NULL) ignore_value(virStringParseYesNo(managed, &def->managed)); + if ((permissive = virXMLPropString(node, "permissive")) != NULL) {+ if ((def->permissive = virTristateBoolTypeFromString(permissive)) < 0) {+ virReportError(VIR_ERR_XML_ERROR, + _("unknown hostdev permissive setting '%s'"), + permissive); + return -1; + } + } + sgio = virXMLPropString(node, "sgio"); rawio = virXMLPropString(node, "rawio"); model = virXMLPropString(node, "model"); @@ -26091,6 +26101,9 @@ virDomainActualNetDefFormat(virBufferPtr buf, virDomainHostdevDefPtr hostdef = virDomainNetGetActualHostdev(def); if (hostdef && hostdef->managed) virBufferAddLit(buf, " managed='yes'"); + if (hostdef && hostdef->permissive) + virBufferAsprintf(buf, " permissive='%s'", + virTristateBoolTypeToString(hostdef->permissive)); } if (def->trustGuestRxFilters) virBufferAsprintf(buf, " trustGuestRxFilters='%s'", @@ -26279,6 +26292,9 @@ virDomainNetDefFormat(virBufferPtr buf, virBufferAsprintf(buf, "<interface type='%s'", typeStr); if (hostdef && hostdef->managed) virBufferAddLit(buf, " managed='yes'"); + if (hostdef && hostdef->permissive) + virBufferAsprintf(buf, " permissive='%s'", + virTristateBoolTypeToString(hostdef->permissive)); if (def->trustGuestRxFilters) virBufferAsprintf(buf, " trustGuestRxFilters='%s'", virTristateBoolTypeToString(def->trustGuestRxFilters)); @@ -28063,6 +28079,9 @@ virDomainHostdevDefFormat(virBufferPtr buf, if (def->mode == VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) { virBufferAsprintf(buf, " managed='%s'", def->managed ? "yes" : "no"); + if (def->permissive) + virBufferAsprintf(buf, " permissive='%s'", + virTristateBoolTypeToString(def->permissive));if (def->source.subsys.type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI &&scsisrc->sgio) diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h index ecb80ef..24ec03c 100644 --- a/src/conf/domain_conf.h +++ b/src/conf/domain_conf.h @@ -345,6 +345,7 @@ struct _virDomainHostdevDef { bool missing; bool readonly; bool shareable; + virTristateBool permissive; union { virDomainHostdevSubsys subsys; virDomainHostdevCaps caps; diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c index 458dfc2..23dd0e4 100644 --- a/src/libxl/libxl_conf.c +++ b/src/libxl/libxl_conf.c@@ -2284,6 +2284,7 @@ libxlMakePCI(virDomainHostdevDefPtr hostdev, libxl_device_pci *pcidev)pcidev->bus = pcisrc->addr.bus; pcidev->dev = pcisrc->addr.slot; pcidev->func = pcisrc->addr.function; + pcidev->permissive = hostdev->permissive == VIR_TRISTATE_BOOL_YES; return 0; }diff --git a/tests/libxlxml2domconfigdata/moredevs-hvm.json b/tests/libxlxml2domconfigdata/moredevs-hvm.jsonindex 7bfd68b..474aa2c 100644 --- a/tests/libxlxml2domconfigdata/moredevs-hvm.json +++ b/tests/libxlxml2domconfigdata/moredevs-hvm.json @@ -88,6 +88,12 @@ "dev": 16, "bus": 10, "rdm_policy": "invalid" + }, + { + "dev": 8, + "bus": 10, + "permissive": true, + "rdm_policy": "invalid" } ], "vfbs": [diff --git a/tests/libxlxml2domconfigdata/moredevs-hvm.xml b/tests/libxlxml2domconfigdata/moredevs-hvm.xmlindex f7eb09f..1b6b738 100644 --- a/tests/libxlxml2domconfigdata/moredevs-hvm.xml +++ b/tests/libxlxml2domconfigdata/moredevs-hvm.xml @@ -48,6 +48,11 @@<address type='pci' domain='0x0000' bus='0x0a' slot='0x10' function='0x0'/></source> </interface> + <hostdev mode='subsystem' type='pci' managed='yes' permissive='yes'> + <source> + <address domain='0x0000' bus='0x0a' slot='0x08' function='0x0'/> + </source> + </hostdev> <graphics type='vnc'/> <video> <model type='cirrus' vram='8192' heads='1' primary='yes'/> base-commit: c3ace7e234ccd43d5a008d93e122e6d47cd58e17
