From: Alex Williamson <alex.williamson@redhat.com> To: Fred Gao <fred.gao@intel.com> Cc: kvm@vger.kernel.org, intel-gfx@lists.freedesktop.org, Zhenyu Wang <zhenyuw@linux.intel.com>, Swee Yee Fonn <swee.yee.fonn@intel.com> Subject: Re: [PATCH v4] vfio/pci: Add support for opregion v2.1+ Date: Fri, 19 Mar 2021 13:26:34 -0600 [thread overview] Message-ID: <20210319132634.5af398b9@omen.home.shazbot.org> (raw) In-Reply-To: <20210302130220.9349-1-fred.gao@intel.com> On Tue, 2 Mar 2021 21:02:20 +0800 Fred Gao <fred.gao@intel.com> wrote: > Before opregion version 2.0 VBT data is stored in opregion mailbox #4, > However, When VBT data exceeds 6KB size and cannot be within mailbox #4 > starting from opregion v2.0+, Extended VBT region, next to opregion, is > used to hold the VBT data, so the total size will be opregion size plus > extended VBT region size. > > since opregion v2.0 with physical host VBT address should not be > practically available for end user, it is not supported. > > Cc: Zhenyu Wang <zhenyuw@linux.intel.com> > Signed-off-by: Swee Yee Fonn <swee.yee.fonn@intel.com> > Signed-off-by: Fred Gao <fred.gao@intel.com> > --- > drivers/vfio/pci/vfio_pci_igd.c | 49 +++++++++++++++++++++++++++++++++ > 1 file changed, 49 insertions(+) > > diff --git a/drivers/vfio/pci/vfio_pci_igd.c b/drivers/vfio/pci/vfio_pci_igd.c > index 53d97f459252..4edb8afcdbfc 100644 > --- a/drivers/vfio/pci/vfio_pci_igd.c > +++ b/drivers/vfio/pci/vfio_pci_igd.c > @@ -21,6 +21,10 @@ > #define OPREGION_SIZE (8 * 1024) > #define OPREGION_PCI_ADDR 0xfc > > +#define OPREGION_RVDA 0x3ba > +#define OPREGION_RVDS 0x3c2 > +#define OPREGION_VERSION 0x16 > + > static size_t vfio_pci_igd_rw(struct vfio_pci_device *vdev, char __user *buf, > size_t count, loff_t *ppos, bool iswrite) > { > @@ -58,6 +62,7 @@ static int vfio_pci_igd_opregion_init(struct vfio_pci_device *vdev) > u32 addr, size; > void *base; > int ret; > + u16 version; > > ret = pci_read_config_dword(vdev->pdev, OPREGION_PCI_ADDR, &addr); > if (ret) > @@ -83,6 +88,50 @@ static int vfio_pci_igd_opregion_init(struct vfio_pci_device *vdev) > > size *= 1024; /* In KB */ > > + /* > + * Support opregion v2.1+ > + * When VBT data exceeds 6KB size and cannot be within mailbox #4 s/#4/#4, then the/ > + * Extended VBT region, next to opregion, is used to hold the VBT data. > + * RVDA (Relative Address of VBT Data from Opregion Base) and RVDS > + * (VBT Data Size) from opregion structure member are used to hold the > + * address from region base and size of VBT data while RVDA/RVDS > + * are not defined before opregion 2.0. > + * > + * opregion 2.0: rvda is the physical VBT address. Let's expand the comment to include why this is a problem to support (virtualization of this register would be required in userspace) and why we're choosing not to manipulate this into a 2.1+ table, which I think is both the practical lack of v2.0 tables in use and any implicit dependencies software may have on the OpRegion version. > + * > + * opregion 2.1+: rvda is unsigned, relative offset from > + * opregion base, and should never point within opregion. And for our purposes must exactly follow the base opregion to avoid exposing unknown host memory to userspace, ie. provide a more descriptive justification for the 2nd error condition below. > + */ > + version = le16_to_cpu(*(__le16 *)(base + OPREGION_VERSION)); > + if (version >= 0x0200) { > + u64 rvda; > + u32 rvds; > + > + rvda = le64_to_cpu(*(__le64 *)(base + OPREGION_RVDA)); > + rvds = le32_to_cpu(*(__le32 *)(base + OPREGION_RVDS)); > + if (rvda && rvds) { > + /* no support for opregion v2.0 with physical VBT address */ > + if (version == 0x0200) { > + memunmap(base); > + pci_err(vdev->pdev, > + "IGD passthrough does not support opregion\n" > + "version 0x%x with physical rvda 0x%llx\n", version, rvda); Why do we need a new line midway through this log message? s/passthrough/assignment/ In testing the version you include the leading zero, do you also want that leading zero in the printed version, ie. %04x? If we get to this code, we already know that both rvda and rvds are non-zero, why is it useful to print the rvda value in this error message? For example, we could print: "IGD assignment does not support opregion version 0x%04x with an extended VBT region" > + return -EINVAL; > + } > + > + if ((u32)rvda != size) { What allows us to assume rvda is a 32bit value given that it's a 64bit register? It seems safer not to include this cast. > + memunmap(base); > + pci_err(vdev->pdev, > + "Extended VBT does not follow opregion !\n" > + "opregion version 0x%x:rvda 0x%llx\n", version, rvda); Again I'm not sure about the usefulness of printing the rvda value on its own. Without knowing the size value it seems meaningless. Like above, get rid of the mid-error new line and random space if you keep the exclamation point. > + return -EINVAL; > + } > + > + /* region size for opregion v2.0+: opregion and VBT size */ > + size += rvds; RVDS is defined as size in bytes, not in kilobytes like the base opregion size, right? Let's include that clarification in the comment since the spec is private. Thanks, Alex > + } > + } > + > if (size != OPREGION_SIZE) { > memunmap(base); > base = memremap(addr, size, MEMREMAP_WB);
WARNING: multiple messages have this Message-ID (diff)
From: Alex Williamson <alex.williamson@redhat.com> To: Fred Gao <fred.gao@intel.com> Cc: Swee Yee Fonn <swee.yee.fonn@intel.com>, intel-gfx@lists.freedesktop.org, kvm@vger.kernel.org Subject: Re: [Intel-gfx] [PATCH v4] vfio/pci: Add support for opregion v2.1+ Date: Fri, 19 Mar 2021 13:26:34 -0600 [thread overview] Message-ID: <20210319132634.5af398b9@omen.home.shazbot.org> (raw) In-Reply-To: <20210302130220.9349-1-fred.gao@intel.com> On Tue, 2 Mar 2021 21:02:20 +0800 Fred Gao <fred.gao@intel.com> wrote: > Before opregion version 2.0 VBT data is stored in opregion mailbox #4, > However, When VBT data exceeds 6KB size and cannot be within mailbox #4 > starting from opregion v2.0+, Extended VBT region, next to opregion, is > used to hold the VBT data, so the total size will be opregion size plus > extended VBT region size. > > since opregion v2.0 with physical host VBT address should not be > practically available for end user, it is not supported. > > Cc: Zhenyu Wang <zhenyuw@linux.intel.com> > Signed-off-by: Swee Yee Fonn <swee.yee.fonn@intel.com> > Signed-off-by: Fred Gao <fred.gao@intel.com> > --- > drivers/vfio/pci/vfio_pci_igd.c | 49 +++++++++++++++++++++++++++++++++ > 1 file changed, 49 insertions(+) > > diff --git a/drivers/vfio/pci/vfio_pci_igd.c b/drivers/vfio/pci/vfio_pci_igd.c > index 53d97f459252..4edb8afcdbfc 100644 > --- a/drivers/vfio/pci/vfio_pci_igd.c > +++ b/drivers/vfio/pci/vfio_pci_igd.c > @@ -21,6 +21,10 @@ > #define OPREGION_SIZE (8 * 1024) > #define OPREGION_PCI_ADDR 0xfc > > +#define OPREGION_RVDA 0x3ba > +#define OPREGION_RVDS 0x3c2 > +#define OPREGION_VERSION 0x16 > + > static size_t vfio_pci_igd_rw(struct vfio_pci_device *vdev, char __user *buf, > size_t count, loff_t *ppos, bool iswrite) > { > @@ -58,6 +62,7 @@ static int vfio_pci_igd_opregion_init(struct vfio_pci_device *vdev) > u32 addr, size; > void *base; > int ret; > + u16 version; > > ret = pci_read_config_dword(vdev->pdev, OPREGION_PCI_ADDR, &addr); > if (ret) > @@ -83,6 +88,50 @@ static int vfio_pci_igd_opregion_init(struct vfio_pci_device *vdev) > > size *= 1024; /* In KB */ > > + /* > + * Support opregion v2.1+ > + * When VBT data exceeds 6KB size and cannot be within mailbox #4 s/#4/#4, then the/ > + * Extended VBT region, next to opregion, is used to hold the VBT data. > + * RVDA (Relative Address of VBT Data from Opregion Base) and RVDS > + * (VBT Data Size) from opregion structure member are used to hold the > + * address from region base and size of VBT data while RVDA/RVDS > + * are not defined before opregion 2.0. > + * > + * opregion 2.0: rvda is the physical VBT address. Let's expand the comment to include why this is a problem to support (virtualization of this register would be required in userspace) and why we're choosing not to manipulate this into a 2.1+ table, which I think is both the practical lack of v2.0 tables in use and any implicit dependencies software may have on the OpRegion version. > + * > + * opregion 2.1+: rvda is unsigned, relative offset from > + * opregion base, and should never point within opregion. And for our purposes must exactly follow the base opregion to avoid exposing unknown host memory to userspace, ie. provide a more descriptive justification for the 2nd error condition below. > + */ > + version = le16_to_cpu(*(__le16 *)(base + OPREGION_VERSION)); > + if (version >= 0x0200) { > + u64 rvda; > + u32 rvds; > + > + rvda = le64_to_cpu(*(__le64 *)(base + OPREGION_RVDA)); > + rvds = le32_to_cpu(*(__le32 *)(base + OPREGION_RVDS)); > + if (rvda && rvds) { > + /* no support for opregion v2.0 with physical VBT address */ > + if (version == 0x0200) { > + memunmap(base); > + pci_err(vdev->pdev, > + "IGD passthrough does not support opregion\n" > + "version 0x%x with physical rvda 0x%llx\n", version, rvda); Why do we need a new line midway through this log message? s/passthrough/assignment/ In testing the version you include the leading zero, do you also want that leading zero in the printed version, ie. %04x? If we get to this code, we already know that both rvda and rvds are non-zero, why is it useful to print the rvda value in this error message? For example, we could print: "IGD assignment does not support opregion version 0x%04x with an extended VBT region" > + return -EINVAL; > + } > + > + if ((u32)rvda != size) { What allows us to assume rvda is a 32bit value given that it's a 64bit register? It seems safer not to include this cast. > + memunmap(base); > + pci_err(vdev->pdev, > + "Extended VBT does not follow opregion !\n" > + "opregion version 0x%x:rvda 0x%llx\n", version, rvda); Again I'm not sure about the usefulness of printing the rvda value on its own. Without knowing the size value it seems meaningless. Like above, get rid of the mid-error new line and random space if you keep the exclamation point. > + return -EINVAL; > + } > + > + /* region size for opregion v2.0+: opregion and VBT size */ > + size += rvds; RVDS is defined as size in bytes, not in kilobytes like the base opregion size, right? Let's include that clarification in the comment since the spec is private. Thanks, Alex > + } > + } > + > if (size != OPREGION_SIZE) { > memunmap(base); > base = memremap(addr, size, MEMREMAP_WB); _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
next prev parent reply other threads:[~2021-03-19 19:27 UTC|newest] Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-12-02 17:12 [PATCH v1] vfio/pci: Add support for opregion v2.0+ Fred Gao 2020-12-02 17:12 ` [Intel-gfx] " Fred Gao 2020-12-02 10:04 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for " Patchwork 2020-12-02 10:35 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork 2020-12-02 12:06 ` [Intel-gfx] ✓ Fi.CI.IGT: " Patchwork 2020-12-02 18:57 ` [PATCH v1] " Alex Williamson 2020-12-02 18:57 ` [Intel-gfx] " Alex Williamson 2020-12-03 9:21 ` Gao, Fred 2020-12-03 9:21 ` [Intel-gfx] " Gao, Fred 2020-12-03 23:38 ` Alex Williamson 2020-12-03 23:38 ` [Intel-gfx] " Alex Williamson 2021-01-18 5:15 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for vfio/pci: Add support for opregion v2.0+ (rev2) Patchwork 2021-01-18 5:46 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork 2021-01-18 12:38 ` [PATCH v2] vfio/pci: Add support for opregion v2.0+ Fred Gao 2021-01-18 12:38 ` [Intel-gfx] " Fred Gao 2021-01-21 20:33 ` Alex Williamson 2021-01-21 20:33 ` [Intel-gfx] " Alex Williamson 2021-02-02 5:09 ` Zhenyu Wang 2021-02-02 5:09 ` [Intel-gfx] " Zhenyu Wang 2021-02-08 17:02 ` [PATCH v3] vfio/pci: Add support for opregion v2.1+ Fred Gao 2021-02-08 17:02 ` [Intel-gfx] " Fred Gao 2021-03-02 13:02 ` [PATCH v4] " Fred Gao 2021-03-02 13:02 ` [Intel-gfx] " Fred Gao 2021-03-19 19:26 ` Alex Williamson [this message] 2021-03-19 19:26 ` Alex Williamson 2021-03-25 8:50 ` Gao, Fred 2021-03-25 8:50 ` [Intel-gfx] " Gao, Fred 2021-03-25 17:09 ` [PATCH v5] " Fred Gao 2021-03-25 17:09 ` [Intel-gfx] " Fred Gao 2021-03-30 9:08 ` Zhenyu Wang 2021-03-30 9:08 ` [Intel-gfx] " Zhenyu Wang 2021-04-06 19:37 ` Alex Williamson 2021-04-06 19:37 ` [Intel-gfx] " Alex Williamson 2021-02-08 9:22 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for vfio/pci: Add support for opregion v2.0+ (rev3) Patchwork 2021-02-08 9:52 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork 2021-03-02 6:49 ` [Intel-gfx] ✗ Fi.CI.BAT: failure for vfio/pci: Add support for opregion v2.0+ (rev4) Patchwork 2021-03-02 12:47 ` Gao, Fred 2021-03-02 6:55 ` [Intel-gfx] ✓ Fi.CI.IGT: success " Patchwork 2021-03-02 17:09 ` [Intel-gfx] ✓ Fi.CI.BAT: " Patchwork 2021-03-25 21:05 ` [Intel-gfx] ✗ Fi.CI.DOCS: warning for vfio/pci: Add support for opregion v2.0+ (rev5) Patchwork 2021-03-29 7:12 ` Gao, Fred 2021-03-29 7:12 ` Zhenyu Wang 2021-03-29 16:19 ` Vudum, Lakshminarayana 2021-03-29 16:22 ` Gao, Fred 2021-03-29 16:35 ` Vudum, Lakshminarayana 2021-03-25 21:31 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork 2021-03-29 15:51 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork 2021-03-29 18:35 ` [Intel-gfx] ✓ Fi.CI.IGT: " Patchwork
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20210319132634.5af398b9@omen.home.shazbot.org \ --to=alex.williamson@redhat.com \ --cc=fred.gao@intel.com \ --cc=intel-gfx@lists.freedesktop.org \ --cc=kvm@vger.kernel.org \ --cc=swee.yee.fonn@intel.com \ --cc=zhenyuw@linux.intel.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.