From: Oleksandr Tyshchenko <[email protected]>
Hello all.
The purpose of this RFC patch series is to add support for restricting memory access under Xen using specific
grant table based DMA ops layer. Patch series is based on Juergen Gross’ initial work [1] which implies using
grant references instead of raw guest physical addresses (GPA) for the virtio communications (some kind of
the software IOMMU).
The high level idea is to create new Xen’s grant table based DMA ops layer for the guest Linux whose main
purpose is to provide a special 64-bit DMA address which is formed by using the grant reference (for a page
to be shared with the backend) with offset and setting the highest address bit (this is for the backend to
be able to distinguish grant ref based DMA address from normal GPA). For this to work we need the ability
to allocate contiguous (consecutive) grant references for multi-page allocations. And the backend then needs
to offer VIRTIO_F_ACCESS_PLATFORM and VIRTIO_F_VERSION_1 feature bits (it must support virtio-mmio modern
transport for 64-bit addresses in the virtqueue).
Xen's grant mapping mechanism is the secure and safe solution to share pages between domains which proven
to work and works for years (in the context of traditional Xen PV drivers for example). So far, the foreign
mapping is used for the virtio backend to map and access guest memory. With the foreign mapping, the backend
is able to map arbitrary pages from the guest memory (or even from Dom0 memory). And as the result, the malicious
backend which runs in a non-trusted domain can take advantage of this. Instead, with the grant mapping
the backend is only allowed to map pages which were explicitly granted by the guest before and nothing else.
According to the discussions in various mainline threads this solution would likely be welcome because it
perfectly fits in the security model Xen provides.
What is more, the grant table based solution requires zero changes to the Xen hypervisor itself at least
with virtio-mmio and DT (in comparison, for example, with "foreign mapping + virtio-iommu" solution which would
require the whole new complex emulator in hypervisor in addition to new functionality/hypercall to pass IOVA
from the virtio backend running elsewhere to the hypervisor and translate it to the GPA before mapping into
P2M or denying the foreign mapping request if no corresponding IOVA-GPA mapping present in the IOMMU page table
for that particular device). We only need to update toolstack to insert a new "xen,dev-domid" property to
the virtio-mmio device node when creating a guest device-tree (this is an indicator for the guest to use grants
and the ID of Xen domain where the corresponding backend resides, it is used as an argument to the grant mapping
APIs). It worth mentioning that toolstack patch is based on non upstreamed yet “Virtio support for toolstack
on Arm” series which is on review now [2].
Please note the following:
- Patch series only covers Arm and virtio-mmio (device-tree) for now. To enable the restricted memory access
feature on Arm the following options should be set:
CONFIG_XEN_VIRTIO = y
CONFIG_XEN_HVM_VIRTIO_GRANT = y
- Some callbacks in xen-virtio DMA ops layer (map_sg/unmap_sg, etc) are not implemented yet as they are not
needed/used in the first prototype
Patch series is rebased on Linux 5.18-rc2 tag and tested on Renesas Salvator-X board + H3 ES3.0 SoC (Arm64)
with standalone userspace (non-Qemu) virtio-mmio based virtio-disk backend running in Driver domain and Linux
guest running on existing virtio-blk driver (frontend). No issues were observed. Guest domain 'reboot/destroy'
use-cases work properly. I have also tested other use-cases such as assigning several virtio block devices
or a mix of virtio and Xen PV block devices to the guest.
1. Xen changes located at (last patch):
https://github.com/otyshchenko1/xen/commits/libxl_virtio_next
2. Linux changes located at:
https://github.com/otyshchenko1/linux/commits/virtio_grant5
3. virtio-disk changes located at:
https://github.com/otyshchenko1/virtio-disk/commits/virtio_grant
Any feedback/help would be highly appreciated.
[1] https://www.youtube.com/watch?v=IrlEdaIUDPk
[2] https://lore.kernel.org/xen-devel/[email protected]/
Juergen Gross (2):
xen/grants: support allocating consecutive grants
virtio: add option to restrict memory access under Xen
Oleksandr Tyshchenko (4):
dt-bindings: xen: Add xen,dev-domid property description for
xen-virtio layer
virtio: Various updates to xen-virtio DMA ops layer
arm/xen: Introduce xen_setup_dma_ops()
arm/xen: Assign xen-virtio DMA ops for virtio devices in Xen guests
.../devicetree/bindings/virtio/xen,dev-domid.yaml | 39 +++
arch/arm/include/asm/xen/xen-ops.h | 1 +
arch/arm/mm/dma-mapping.c | 5 +-
arch/arm/xen/enlighten.c | 11 +
arch/arm64/include/asm/xen/xen-ops.h | 1 +
arch/arm64/mm/dma-mapping.c | 5 +-
arch/x86/mm/init.c | 15 +
arch/x86/mm/mem_encrypt.c | 5 -
arch/x86/xen/Kconfig | 9 +
drivers/xen/Kconfig | 20 ++
drivers/xen/Makefile | 1 +
drivers/xen/grant-table.c | 238 +++++++++++++--
drivers/xen/xen-virtio.c | 335 +++++++++++++++++++++
include/xen/arm/xen-ops.h | 20 ++
include/xen/grant_table.h | 4 +
include/xen/xen-ops.h | 13 +
16 files changed, 679 insertions(+), 43 deletions(-)
create mode 100644 Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml
create mode 100644 arch/arm/include/asm/xen/xen-ops.h
create mode 100644 arch/arm64/include/asm/xen/xen-ops.h
create mode 100644 drivers/xen/xen-virtio.c
create mode 100644 include/xen/arm/xen-ops.h
--
2.7.4
From: Juergen Gross <[email protected]>
In order to support virtio in Xen guests add a config option enabling
the user to specify whether in all Xen guests virtio should be able to
access memory via Xen grant mappings only on the host side.
This applies to fully virtualized guests only, as for paravirtualized
guests this is mandatory.
This requires to switch arch_has_restricted_virtio_memory_access()
from a pure stub to a real function on x86 systems (Arm systems are
not covered by now).
Add the needed functionality by providing a special set of DMA ops
handling the needed grant operations for the I/O pages.
Signed-off-by: Juergen Gross <[email protected]>
---
arch/x86/mm/init.c | 15 ++++
arch/x86/mm/mem_encrypt.c | 5 --
arch/x86/xen/Kconfig | 9 +++
drivers/xen/Kconfig | 20 ++++++
drivers/xen/Makefile | 1 +
drivers/xen/xen-virtio.c | 177 ++++++++++++++++++++++++++++++++++++++++++++++
include/xen/xen-ops.h | 8 +++
7 files changed, 230 insertions(+), 5 deletions(-)
create mode 100644 drivers/xen/xen-virtio.c
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index d8cfce2..526a3b2 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -8,6 +8,8 @@
#include <linux/kmemleak.h>
#include <linux/sched/task.h>
+#include <xen/xen.h>
+
#include <asm/set_memory.h>
#include <asm/e820/api.h>
#include <asm/init.h>
@@ -1065,3 +1067,16 @@ unsigned long max_swapfile_size(void)
return pages;
}
#endif
+
+#ifdef CONFIG_ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
+int arch_has_restricted_virtio_memory_access(void)
+{
+ if (IS_ENABLED(CONFIG_XEN_PV_VIRTIO) && xen_pv_domain())
+ return 1;
+ if (IS_ENABLED(CONFIG_XEN_HVM_VIRTIO_GRANT) && xen_hvm_domain())
+ return 1;
+
+ return cc_platform_has(CC_ATTR_GUEST_MEM_ENCRYPT);
+}
+EXPORT_SYMBOL_GPL(arch_has_restricted_virtio_memory_access);
+#endif
diff --git a/arch/x86/mm/mem_encrypt.c b/arch/x86/mm/mem_encrypt.c
index 50d2099..dda020f 100644
--- a/arch/x86/mm/mem_encrypt.c
+++ b/arch/x86/mm/mem_encrypt.c
@@ -77,8 +77,3 @@ void __init mem_encrypt_init(void)
print_mem_encrypt_feature_info();
}
-int arch_has_restricted_virtio_memory_access(void)
-{
- return cc_platform_has(CC_ATTR_GUEST_MEM_ENCRYPT);
-}
-EXPORT_SYMBOL_GPL(arch_has_restricted_virtio_memory_access);
diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig
index 85246dd..dffdffd 100644
--- a/arch/x86/xen/Kconfig
+++ b/arch/x86/xen/Kconfig
@@ -92,3 +92,12 @@ config XEN_DOM0
select X86_X2APIC if XEN_PVH && X86_64
help
Support running as a Xen Dom0 guest.
+
+config XEN_PV_VIRTIO
+ bool "Xen virtio support for PV guests"
+ depends on XEN_VIRTIO && XEN_PV
+ default y
+ help
+ Support virtio for running as a paravirtualized guest. This will
+ need support on the backend side (qemu or kernel, depending on the
+ virtio device types used).
diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig
index 120d32f..fc61f7a 100644
--- a/drivers/xen/Kconfig
+++ b/drivers/xen/Kconfig
@@ -335,4 +335,24 @@ config XEN_UNPOPULATED_ALLOC
having to balloon out RAM regions in order to obtain physical memory
space to create such mappings.
+config XEN_VIRTIO
+ bool "Xen virtio support"
+ default n
+ depends on VIRTIO && DMA_OPS
+ select ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
+ help
+ Enable virtio support for running as Xen guest. Depending on the
+ guest type this will require special support on the backend side
+ (qemu or kernel, depending on the virtio device types used).
+
+config XEN_HVM_VIRTIO_GRANT
+ bool "Require virtio for fully virtualized guests to use grant mappings"
+ depends on XEN_VIRTIO && X86_64
+ default y
+ help
+ Require virtio for fully virtualized guests to use grant mappings.
+ This will avoid the need to give the backend the right to map all
+ of the guest memory. This will need support on the backend side
+ (qemu or kernel, depending on the virtio device types used).
+
endmenu
diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile
index 5aae66e..767009c 100644
--- a/drivers/xen/Makefile
+++ b/drivers/xen/Makefile
@@ -39,3 +39,4 @@ xen-gntalloc-y := gntalloc.o
xen-privcmd-y := privcmd.o privcmd-buf.o
obj-$(CONFIG_XEN_FRONT_PGDIR_SHBUF) += xen-front-pgdir-shbuf.o
obj-$(CONFIG_XEN_UNPOPULATED_ALLOC) += unpopulated-alloc.o
+obj-$(CONFIG_XEN_VIRTIO) += xen-virtio.o
diff --git a/drivers/xen/xen-virtio.c b/drivers/xen/xen-virtio.c
new file mode 100644
index 00000000..cfd5eda
--- /dev/null
+++ b/drivers/xen/xen-virtio.c
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/******************************************************************************
+ * Xen virtio driver - enables using virtio devices in Xen guests.
+ *
+ * Copyright (c) 2021, Juergen Gross <[email protected]>
+ */
+
+#include <linux/module.h>
+#include <linux/dma-map-ops.h>
+#include <linux/pci.h>
+#include <linux/pfn.h>
+#include <linux/virtio_config.h>
+#include <xen/xen.h>
+#include <xen/grant_table.h>
+
+#define XEN_GRANT_ADDR_OFF 0x8000000000000000ULL
+
+static inline dma_addr_t grant_to_dma(grant_ref_t grant)
+{
+ return XEN_GRANT_ADDR_OFF | ((dma_addr_t)grant << PAGE_SHIFT);
+}
+
+static inline grant_ref_t dma_to_grant(dma_addr_t dma)
+{
+ return (grant_ref_t)((dma & ~XEN_GRANT_ADDR_OFF) >> PAGE_SHIFT);
+}
+
+/*
+ * DMA ops for Xen virtio frontends.
+ *
+ * Used to act as a kind of software IOMMU for Xen guests by using grants as
+ * DMA addresses.
+ * Such a DMA address is formed by using the grant reference as a frame
+ * number and setting the highest address bit (this bit is for the backend
+ * to be able to distinguish it from e.g. a mmio address).
+ *
+ * Note that for now we hard wire dom0 to be the backend domain. In order to
+ * support any domain as backend we'd need to add a way to communicate the
+ * domid of this backend, e.g. via Xenstore or via the PCI-device's config
+ * space.
+ */
+static void *xen_virtio_dma_alloc(struct device *dev, size_t size,
+ dma_addr_t *dma_handle, gfp_t gfp,
+ unsigned long attrs)
+{
+ unsigned int n_pages = PFN_UP(size);
+ unsigned int i;
+ unsigned long pfn;
+ grant_ref_t grant;
+ void *ret;
+
+ ret = (void *)__get_free_pages(gfp, get_order(size));
+ if (!ret)
+ return NULL;
+
+ pfn = virt_to_pfn(ret);
+
+ if (gnttab_alloc_grant_reference_seq(n_pages, &grant)) {
+ free_pages((unsigned long)ret, get_order(size));
+ return NULL;
+ }
+
+ for (i = 0; i < n_pages; i++) {
+ gnttab_grant_foreign_access_ref(grant + i, 0,
+ pfn_to_gfn(pfn + i), 0);
+ }
+
+ *dma_handle = grant_to_dma(grant);
+
+ return ret;
+}
+
+static void xen_virtio_dma_free(struct device *dev, size_t size, void *vaddr,
+ dma_addr_t dma_handle, unsigned long attrs)
+{
+ unsigned int n_pages = PFN_UP(size);
+ unsigned int i;
+ grant_ref_t grant;
+
+ grant = dma_to_grant(dma_handle);
+
+ for (i = 0; i < n_pages; i++)
+ gnttab_end_foreign_access_ref(grant + i);
+
+ gnttab_free_grant_reference_seq(grant, n_pages);
+
+ free_pages((unsigned long)vaddr, get_order(size));
+}
+
+static struct page *xen_virtio_dma_alloc_pages(struct device *dev, size_t size,
+ dma_addr_t *dma_handle,
+ enum dma_data_direction dir,
+ gfp_t gfp)
+{
+ WARN_ONCE(1, "xen_virtio_dma_alloc_pages size %ld\n", size);
+ return NULL;
+}
+
+static void xen_virtio_dma_free_pages(struct device *dev, size_t size,
+ struct page *vaddr, dma_addr_t dma_handle,
+ enum dma_data_direction dir)
+{
+ WARN_ONCE(1, "xen_virtio_dma_free_pages size %ld\n", size);
+}
+
+static dma_addr_t xen_virtio_dma_map_page(struct device *dev, struct page *page,
+ unsigned long offset, size_t size,
+ enum dma_data_direction dir,
+ unsigned long attrs)
+{
+ grant_ref_t grant;
+
+ if (gnttab_alloc_grant_references(1, &grant))
+ return 0;
+
+ gnttab_grant_foreign_access_ref(grant, 0, xen_page_to_gfn(page),
+ dir == DMA_TO_DEVICE);
+
+ return grant_to_dma(grant) + offset;
+}
+
+static void xen_virtio_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
+ size_t size, enum dma_data_direction dir,
+ unsigned long attrs)
+{
+ grant_ref_t grant;
+
+ grant = dma_to_grant(dma_handle);
+
+ gnttab_end_foreign_access_ref(grant);
+
+ gnttab_free_grant_reference(grant);
+}
+
+static int xen_virtio_dma_map_sg(struct device *dev, struct scatterlist *sg,
+ int nents, enum dma_data_direction dir,
+ unsigned long attrs)
+{
+ WARN_ONCE(1, "xen_virtio_dma_map_sg nents %d\n", nents);
+ return -EINVAL;
+}
+
+static void xen_virtio_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
+ int nents, enum dma_data_direction dir,
+ unsigned long attrs)
+{
+ WARN_ONCE(1, "xen_virtio_dma_unmap_sg nents %d\n", nents);
+}
+
+static int xen_virtio_dma_dma_supported(struct device *dev, u64 mask)
+{
+ return 1;
+}
+
+static const struct dma_map_ops xen_virtio_dma_ops = {
+ .alloc = xen_virtio_dma_alloc,
+ .free = xen_virtio_dma_free,
+ .alloc_pages = xen_virtio_dma_alloc_pages,
+ .free_pages = xen_virtio_dma_free_pages,
+ .mmap = dma_common_mmap,
+ .get_sgtable = dma_common_get_sgtable,
+ .map_page = xen_virtio_dma_map_page,
+ .unmap_page = xen_virtio_dma_unmap_page,
+ .map_sg = xen_virtio_dma_map_sg,
+ .unmap_sg = xen_virtio_dma_unmap_sg,
+ .dma_supported = xen_virtio_dma_dma_supported,
+};
+
+void xen_virtio_setup_dma_ops(struct device *dev)
+{
+ dev->dma_ops = &xen_virtio_dma_ops;
+}
+EXPORT_SYMBOL_GPL(xen_virtio_setup_dma_ops);
+
+MODULE_DESCRIPTION("Xen virtio support driver");
+MODULE_AUTHOR("Juergen Gross <[email protected]>");
+MODULE_LICENSE("GPL");
diff --git a/include/xen/xen-ops.h b/include/xen/xen-ops.h
index a3584a3..ae3c1bc 100644
--- a/include/xen/xen-ops.h
+++ b/include/xen/xen-ops.h
@@ -221,4 +221,12 @@ static inline void xen_preemptible_hcall_end(void) { }
#endif /* CONFIG_XEN_PV && !CONFIG_PREEMPTION */
+#ifdef CONFIG_XEN_VIRTIO
+void xen_virtio_setup_dma_ops(struct device *dev);
+#else
+static inline void xen_virtio_setup_dma_ops(struct device *dev)
+{
+}
+#endif /* CONFIG_XEN_VIRTIO */
+
#endif /* INCLUDE_XEN_OPS_H */
--
2.7.4
On 15.04.22 11:44, Michael S. Tsirkin wrote:
Hello Michael
> On Thu, Apr 14, 2022 at 10:19:27PM +0300, Oleksandr Tyshchenko wrote:
>> From: Oleksandr Tyshchenko <[email protected]>
>>
>> Hello all.
>>
>> The purpose of this RFC patch series is to add support for restricting memory access under Xen using specific
>> grant table based DMA ops layer. Patch series is based on Juergen Gross’ initial work [1] which implies using
>> grant references instead of raw guest physical addresses (GPA) for the virtio communications (some kind of
>> the software IOMMU).
>>
>> The high level idea is to create new Xen’s grant table based DMA ops layer for the guest Linux whose main
>> purpose is to provide a special 64-bit DMA address which is formed by using the grant reference (for a page
>> to be shared with the backend) with offset and setting the highest address bit (this is for the backend to
>> be able to distinguish grant ref based DMA address from normal GPA). For this to work we need the ability
>> to allocate contiguous (consecutive) grant references for multi-page allocations. And the backend then needs
>> to offer VIRTIO_F_ACCESS_PLATFORM and VIRTIO_F_VERSION_1 feature bits (it must support virtio-mmio modern
>> transport for 64-bit addresses in the virtqueue).
> I'm not enough of a xen expert to review this, and I didn't get
> all patches, but I'm very happy to see that approach being
> taken. VIRTIO_F_ACCESS_PLATFORM and VIRTIO_F_VERSION_1 are
> exactly the way to declare not all of memory is accessible.
> Thanks!
I am happy to hear that! Thank you.
Regarding the "all patches" I have already redirect missing ones, I hope
you and Christoph will get them.
Sorry for the inconvenience.
>
>> Xen's grant mapping mechanism is the secure and safe solution to share pages between domains which proven
>> to work and works for years (in the context of traditional Xen PV drivers for example). So far, the foreign
>> mapping is used for the virtio backend to map and access guest memory. With the foreign mapping, the backend
>> is able to map arbitrary pages from the guest memory (or even from Dom0 memory). And as the result, the malicious
>> backend which runs in a non-trusted domain can take advantage of this. Instead, with the grant mapping
>> the backend is only allowed to map pages which were explicitly granted by the guest before and nothing else.
>> According to the discussions in various mainline threads this solution would likely be welcome because it
>> perfectly fits in the security model Xen provides.
>>
>> What is more, the grant table based solution requires zero changes to the Xen hypervisor itself at least
>> with virtio-mmio and DT (in comparison, for example, with "foreign mapping + virtio-iommu" solution which would
>> require the whole new complex emulator in hypervisor in addition to new functionality/hypercall to pass IOVA
>> from the virtio backend running elsewhere to the hypervisor and translate it to the GPA before mapping into
>> P2M or denying the foreign mapping request if no corresponding IOVA-GPA mapping present in the IOMMU page table
>> for that particular device). We only need to update toolstack to insert a new "xen,dev-domid" property to
>> the virtio-mmio device node when creating a guest device-tree (this is an indicator for the guest to use grants
>> and the ID of Xen domain where the corresponding backend resides, it is used as an argument to the grant mapping
>> APIs). It worth mentioning that toolstack patch is based on non upstreamed yet “Virtio support for toolstack
>> on Arm” series which is on review now [2].
>>
>> Please note the following:
>> - Patch series only covers Arm and virtio-mmio (device-tree) for now. To enable the restricted memory access
>> feature on Arm the following options should be set:
>> CONFIG_XEN_VIRTIO = y
>> CONFIG_XEN_HVM_VIRTIO_GRANT = y
>> - Some callbacks in xen-virtio DMA ops layer (map_sg/unmap_sg, etc) are not implemented yet as they are not
>> needed/used in the first prototype
>>
>> Patch series is rebased on Linux 5.18-rc2 tag and tested on Renesas Salvator-X board + H3 ES3.0 SoC (Arm64)
>> with standalone userspace (non-Qemu) virtio-mmio based virtio-disk backend running in Driver domain and Linux
>> guest running on existing virtio-blk driver (frontend). No issues were observed. Guest domain 'reboot/destroy'
>> use-cases work properly. I have also tested other use-cases such as assigning several virtio block devices
>> or a mix of virtio and Xen PV block devices to the guest.
>>
>> 1. Xen changes located at (last patch):
>> https://github.com/otyshchenko1/xen/commits/libxl_virtio_next
>> 2. Linux changes located at:
>> https://github.com/otyshchenko1/linux/commits/virtio_grant5
>> 3. virtio-disk changes located at:
>> https://github.com/otyshchenko1/virtio-disk/commits/virtio_grant
>>
>> Any feedback/help would be highly appreciated.
>>
>> [1] https://www.youtube.com/watch?v=IrlEdaIUDPk
>> [2] https://lore.kernel.org/xen-devel/[email protected]/
>>
>> Juergen Gross (2):
>> xen/grants: support allocating consecutive grants
>> virtio: add option to restrict memory access under Xen
>>
>> Oleksandr Tyshchenko (4):
>> dt-bindings: xen: Add xen,dev-domid property description for
>> xen-virtio layer
>> virtio: Various updates to xen-virtio DMA ops layer
>> arm/xen: Introduce xen_setup_dma_ops()
>> arm/xen: Assign xen-virtio DMA ops for virtio devices in Xen guests
>>
>> .../devicetree/bindings/virtio/xen,dev-domid.yaml | 39 +++
>> arch/arm/include/asm/xen/xen-ops.h | 1 +
>> arch/arm/mm/dma-mapping.c | 5 +-
>> arch/arm/xen/enlighten.c | 11 +
>> arch/arm64/include/asm/xen/xen-ops.h | 1 +
>> arch/arm64/mm/dma-mapping.c | 5 +-
>> arch/x86/mm/init.c | 15 +
>> arch/x86/mm/mem_encrypt.c | 5 -
>> arch/x86/xen/Kconfig | 9 +
>> drivers/xen/Kconfig | 20 ++
>> drivers/xen/Makefile | 1 +
>> drivers/xen/grant-table.c | 238 +++++++++++++--
>> drivers/xen/xen-virtio.c | 335 +++++++++++++++++++++
>> include/xen/arm/xen-ops.h | 20 ++
>> include/xen/grant_table.h | 4 +
>> include/xen/xen-ops.h | 13 +
>> 16 files changed, 679 insertions(+), 43 deletions(-)
>> create mode 100644 Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml
>> create mode 100644 arch/arm/include/asm/xen/xen-ops.h
>> create mode 100644 arch/arm64/include/asm/xen/xen-ops.h
>> create mode 100644 drivers/xen/xen-virtio.c
>> create mode 100644 include/xen/arm/xen-ops.h
>>
>> --
>> 2.7.4
--
Regards,
Oleksandr Tyshchenko
From: Oleksandr Tyshchenko <[email protected]>
In the context of current patch do the following:
1. Update code to support virtio-mmio devices
2. Introduce struct xen_virtio_data and account passed virtio devices
(using list) as we need to store some per-device data
3. Add multi-page support for xen_virtio_dma_map(unmap)_page callbacks
4. Harden code against malicious backend
5. Change to use alloc_pages_exact() instead of __get_free_pages()
6. Introduce locking scheme to protect mappings (I am not 100% sure
whether per-device lock is really needed)
7. Handle virtio device's DMA mask
8. Retrieve the ID of backend domain from DT for virtio-mmio device
instead of hardcoding it.
Signed-off-by: Oleksandr Tyshchenko <[email protected]>
---
arch/arm/xen/enlighten.c | 11 +++
drivers/xen/Kconfig | 2 +-
drivers/xen/xen-virtio.c | 200 ++++++++++++++++++++++++++++++++++++++++++-----
include/xen/xen-ops.h | 5 ++
4 files changed, 196 insertions(+), 22 deletions(-)
diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
index ec5b082..870d92f 100644
--- a/arch/arm/xen/enlighten.c
+++ b/arch/arm/xen/enlighten.c
@@ -409,6 +409,17 @@ int __init arch_xen_unpopulated_init(struct resource **res)
}
#endif
+#ifdef CONFIG_ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
+int arch_has_restricted_virtio_memory_access(void)
+{
+ if (IS_ENABLED(CONFIG_XEN_HVM_VIRTIO_GRANT) && xen_hvm_domain())
+ return 1;
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(arch_has_restricted_virtio_memory_access);
+#endif
+
static void __init xen_dt_guest_init(void)
{
struct device_node *xen_node;
diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig
index fc61f7a..56afe6a 100644
--- a/drivers/xen/Kconfig
+++ b/drivers/xen/Kconfig
@@ -347,7 +347,7 @@ config XEN_VIRTIO
config XEN_HVM_VIRTIO_GRANT
bool "Require virtio for fully virtualized guests to use grant mappings"
- depends on XEN_VIRTIO && X86_64
+ depends on XEN_VIRTIO && (X86_64 || ARM || ARM64)
default y
help
Require virtio for fully virtualized guests to use grant mappings.
diff --git a/drivers/xen/xen-virtio.c b/drivers/xen/xen-virtio.c
index cfd5eda..c5b2ec9 100644
--- a/drivers/xen/xen-virtio.c
+++ b/drivers/xen/xen-virtio.c
@@ -7,12 +7,26 @@
#include <linux/module.h>
#include <linux/dma-map-ops.h>
+#include <linux/of.h>
#include <linux/pci.h>
#include <linux/pfn.h>
#include <linux/virtio_config.h>
#include <xen/xen.h>
#include <xen/grant_table.h>
+struct xen_virtio_data {
+ /* The ID of backend domain */
+ domid_t dev_domid;
+ struct device *dev;
+ struct list_head list;
+ spinlock_t lock;
+ /* Is device behaving sane? */
+ bool broken;
+};
+
+static LIST_HEAD(xen_virtio_devices);
+static DEFINE_SPINLOCK(xen_virtio_lock);
+
#define XEN_GRANT_ADDR_OFF 0x8000000000000000ULL
static inline dma_addr_t grant_to_dma(grant_ref_t grant)
@@ -25,6 +39,25 @@ static inline grant_ref_t dma_to_grant(dma_addr_t dma)
return (grant_ref_t)((dma & ~XEN_GRANT_ADDR_OFF) >> PAGE_SHIFT);
}
+static struct xen_virtio_data *find_xen_virtio_data(struct device *dev)
+{
+ struct xen_virtio_data *data = NULL;
+ bool found = false;
+
+ spin_lock(&xen_virtio_lock);
+
+ list_for_each_entry( data, &xen_virtio_devices, list) {
+ if (data->dev == dev) {
+ found = true;
+ break;
+ }
+ }
+
+ spin_unlock(&xen_virtio_lock);
+
+ return found ? data : NULL;
+}
+
/*
* DMA ops for Xen virtio frontends.
*
@@ -43,48 +76,78 @@ static void *xen_virtio_dma_alloc(struct device *dev, size_t size,
dma_addr_t *dma_handle, gfp_t gfp,
unsigned long attrs)
{
- unsigned int n_pages = PFN_UP(size);
- unsigned int i;
+ struct xen_virtio_data *data;
+ unsigned int i, n_pages = PFN_UP(size);
unsigned long pfn;
grant_ref_t grant;
- void *ret;
+ void *ret = NULL;
- ret = (void *)__get_free_pages(gfp, get_order(size));
- if (!ret)
+ data = find_xen_virtio_data(dev);
+ if (!data)
return NULL;
+ spin_lock(&data->lock);
+
+ if (unlikely(data->broken))
+ goto out;
+
+ ret = alloc_pages_exact(n_pages * PAGE_SIZE, gfp);
+ if (!ret)
+ goto out;
+
pfn = virt_to_pfn(ret);
if (gnttab_alloc_grant_reference_seq(n_pages, &grant)) {
- free_pages((unsigned long)ret, get_order(size));
- return NULL;
+ free_pages_exact(ret, n_pages * PAGE_SIZE);
+ ret = NULL;
+ goto out;
}
for (i = 0; i < n_pages; i++) {
- gnttab_grant_foreign_access_ref(grant + i, 0,
+ gnttab_grant_foreign_access_ref(grant + i, data->dev_domid,
pfn_to_gfn(pfn + i), 0);
}
*dma_handle = grant_to_dma(grant);
+out:
+ spin_unlock(&data->lock);
+
return ret;
}
static void xen_virtio_dma_free(struct device *dev, size_t size, void *vaddr,
dma_addr_t dma_handle, unsigned long attrs)
{
- unsigned int n_pages = PFN_UP(size);
- unsigned int i;
+ struct xen_virtio_data *data;
+ unsigned int i, n_pages = PFN_UP(size);
grant_ref_t grant;
+ data = find_xen_virtio_data(dev);
+ if (!data)
+ return;
+
+ spin_lock(&data->lock);
+
+ if (unlikely(data->broken))
+ goto out;
+
grant = dma_to_grant(dma_handle);
- for (i = 0; i < n_pages; i++)
- gnttab_end_foreign_access_ref(grant + i);
+ for (i = 0; i < n_pages; i++) {
+ if (unlikely(!gnttab_end_foreign_access_ref(grant + i))) {
+ dev_alert(dev, "Grant still in use by backend domain, disabled for further use\n");
+ data->broken = true;
+ goto out;
+ }
+ }
gnttab_free_grant_reference_seq(grant, n_pages);
- free_pages((unsigned long)vaddr, get_order(size));
+ free_pages_exact(vaddr, n_pages * PAGE_SIZE);
+
+out:
+ spin_unlock(&data->lock);
}
static struct page *xen_virtio_dma_alloc_pages(struct device *dev, size_t size,
@@ -108,28 +171,71 @@ static dma_addr_t xen_virtio_dma_map_page(struct device *dev, struct page *page,
enum dma_data_direction dir,
unsigned long attrs)
{
+ struct xen_virtio_data *data;
+ unsigned int i, n_pages = PFN_UP(size);
grant_ref_t grant;
+ dma_addr_t dma_handle = DMA_MAPPING_ERROR;
+
+ BUG_ON(dir == DMA_NONE);
+
+ data = find_xen_virtio_data(dev);
+ if (!data)
+ return DMA_MAPPING_ERROR;
+
+ spin_lock(&data->lock);
- if (gnttab_alloc_grant_references(1, &grant))
- return 0;
+ if (unlikely(data->broken))
+ goto out;
- gnttab_grant_foreign_access_ref(grant, 0, xen_page_to_gfn(page),
- dir == DMA_TO_DEVICE);
+ if (gnttab_alloc_grant_reference_seq(n_pages, &grant))
+ goto out;
- return grant_to_dma(grant) + offset;
+ for (i = 0; i < n_pages; i++) {
+ gnttab_grant_foreign_access_ref(grant + i, data->dev_domid,
+ xen_page_to_gfn(page) + i, dir == DMA_TO_DEVICE);
+ }
+
+ dma_handle = grant_to_dma(grant) + offset;
+
+out:
+ spin_unlock(&data->lock);
+
+ return dma_handle;
}
static void xen_virtio_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
size_t size, enum dma_data_direction dir,
unsigned long attrs)
{
+ struct xen_virtio_data *data;
+ unsigned int i, n_pages = PFN_UP(size);
grant_ref_t grant;
+ BUG_ON(dir == DMA_NONE);
+
+ data = find_xen_virtio_data(dev);
+ if (!data)
+ return;
+
+ spin_lock(&data->lock);
+
+ if (unlikely(data->broken))
+ goto out;
+
grant = dma_to_grant(dma_handle);
- gnttab_end_foreign_access_ref(grant);
+ for (i = 0; i < n_pages; i++) {
+ if (unlikely(!gnttab_end_foreign_access_ref(grant + i))) {
+ dev_alert(dev, "Grant still in use by backend domain, disabled for further use\n");
+ data->broken = true;
+ goto out;
+ }
+ }
+
+ gnttab_free_grant_reference_seq(grant, n_pages);
- gnttab_free_grant_reference(grant);
+out:
+ spin_unlock(&data->lock);
}
static int xen_virtio_dma_map_sg(struct device *dev, struct scatterlist *sg,
@@ -149,7 +255,7 @@ static void xen_virtio_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
static int xen_virtio_dma_dma_supported(struct device *dev, u64 mask)
{
- return 1;
+ return mask == DMA_BIT_MASK(64);
}
static const struct dma_map_ops xen_virtio_dma_ops = {
@@ -166,9 +272,61 @@ static const struct dma_map_ops xen_virtio_dma_ops = {
.dma_supported = xen_virtio_dma_dma_supported,
};
+bool xen_is_virtio_device(struct device *dev)
+{
+ /* XXX Handle only DT devices for now */
+ if (!dev->of_node)
+ return false;
+
+ if (!of_device_is_compatible(dev->of_node, "virtio,mmio"))
+ return false;
+
+ return of_property_read_bool(dev->of_node, "xen,dev-domid");
+}
+EXPORT_SYMBOL_GPL(xen_is_virtio_device);
+
void xen_virtio_setup_dma_ops(struct device *dev)
{
+ struct xen_virtio_data *data;
+ uint32_t dev_domid;
+
+ data = find_xen_virtio_data(dev);
+ if (data) {
+ dev_err(dev, "xen_virtio data is already created\n");
+ return;
+ }
+
+ if (dev_is_pci(dev)) {
+ /* XXX Leave it hard wired to dom0 for now */
+ dev_domid = 0;
+ } else if (dev->of_node) {
+ if (of_property_read_u32(dev->of_node, "xen,dev-domid", &dev_domid)) {
+ dev_err(dev, "xen,dev-domid property is not present\n");
+ goto err;
+ }
+ } else
+ /* The ACPI case is not supported */
+ goto err;
+
+ data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
+ if (!data) {
+ dev_err(dev, "Сannot allocate xen_virtio data\n");
+ goto err;
+ }
+ data->dev_domid = dev_domid;
+ data->dev = dev;
+ spin_lock_init(&data->lock);
+
+ spin_lock(&xen_virtio_lock);
+ list_add(&data->list, &xen_virtio_devices);
+ spin_unlock(&xen_virtio_lock);
+
dev->dma_ops = &xen_virtio_dma_ops;
+
+ return;
+
+err:
+ dev_err(dev, "Сannot set up xen_virtio DMA ops, retain platform DMA ops\n");
}
EXPORT_SYMBOL_GPL(xen_virtio_setup_dma_ops);
diff --git a/include/xen/xen-ops.h b/include/xen/xen-ops.h
index ae3c1bc..fdbcb99 100644
--- a/include/xen/xen-ops.h
+++ b/include/xen/xen-ops.h
@@ -223,10 +223,15 @@ static inline void xen_preemptible_hcall_end(void) { }
#ifdef CONFIG_XEN_VIRTIO
void xen_virtio_setup_dma_ops(struct device *dev);
+bool xen_is_virtio_device(struct device *dev);
#else
static inline void xen_virtio_setup_dma_ops(struct device *dev)
{
}
+static inline bool xen_is_virtio_device(struct device *dev)
+{
+ return false;
+}
#endif /* CONFIG_XEN_VIRTIO */
#endif /* INCLUDE_XEN_OPS_H */
--
2.7.4
On 15.04.22 10:41, Christoph Hellwig wrote:
Hello Christoph
> I can only see three out of 6 patches on the linux-arm-kernel list,
> which makes reviewing this impossible.
Oops, I will add linux-arm-kernel. I blindly followed what
get_maintainer.pl suggested for each patch plus added manually some Xen
folks,
but, indeed, the first three patches add the base of this enabling work.
> Also please Cc me directly
> on any series doing crazy things with dma ops. Thanks!
yes, sure.
--
Regards,
Oleksandr Tyshchenko
From: Oleksandr Tyshchenko <[email protected]>
Call xen_virtio_setup_dma_ops() only for Xen-aware virtio devices
in Xen guests if restricted access to the guest memory is enabled.
Signed-off-by: Oleksandr Tyshchenko <[email protected]>
---
include/xen/arm/xen-ops.h | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/include/xen/arm/xen-ops.h b/include/xen/arm/xen-ops.h
index 621da05..28b2ad3 100644
--- a/include/xen/arm/xen-ops.h
+++ b/include/xen/arm/xen-ops.h
@@ -2,12 +2,19 @@
#ifndef _ASM_ARM_XEN_OPS_H
#define _ASM_ARM_XEN_OPS_H
+#include <linux/virtio_config.h>
#include <xen/swiotlb-xen.h>
+#include <xen/xen-ops.h>
static inline void xen_setup_dma_ops(struct device *dev)
{
if (xen_swiotlb_detect())
dev->dma_ops = &xen_swiotlb_dma_ops;
+
+#ifdef CONFIG_XEN_VIRTIO
+ if (arch_has_restricted_virtio_memory_access() && xen_is_virtio_device(dev))
+ xen_virtio_setup_dma_ops(dev);
+#endif
}
#endif /* _ASM_ARM_XEN_OPS_H */
--
2.7.4
I can only see three out of 6 patches on the linux-arm-kernel list,
which makes reviewing this impossible. Also please Cc me directly
on any series doing crazy things with dma ops. Thanks!
From: Oleksandr Tyshchenko <[email protected]>
Introduce Xen specific binding for the virtio-mmio device to be used
by Xen virtio support driver in a subsequent commit.
This binding specifies the ID of Xen domain where the corresponding
device (backend) resides. This is needed for the option to restrict
memory access using Xen grant mappings to work.
Signed-off-by: Oleksandr Tyshchenko <[email protected]>
---
.../devicetree/bindings/virtio/xen,dev-domid.yaml | 39 ++++++++++++++++++++++
1 file changed, 39 insertions(+)
create mode 100644 Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml
diff --git a/Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml b/Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml
new file mode 100644
index 00000000..78be993
--- /dev/null
+++ b/Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml
@@ -0,0 +1,39 @@
+# SPDX-License-Identifier: (GPL-2.0-only or BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/virtio/xen,dev-domid.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Xen specific binding for the virtio device
+
+maintainers:
+ - Oleksandr Tyshchenko <[email protected]>
+
+select: true
+
+description:
+ This binding specifies the ID of Xen domain where the corresponding device
+ (backend) resides. This is needed for the option to restrict memory access
+ using Xen grant mappings to work.
+
+ Note that current and generic "iommus" bindings are mutually exclusive, since
+ the restricted memory access model on Xen behaves as a kind of software IOMMU.
+
+properties:
+ xen,dev-domid:
+ $ref: /schemas/types.yaml#/definitions/uint32
+ description:
+ Should contain the ID of device's domain.
+
+additionalProperties: true
+
+examples:
+ - |
+ virtio_block@3000 {
+ compatible = "virtio,mmio";
+ reg = <0x3000 0x100>;
+ interrupts = <41>;
+
+ /* The device is located in Xen domain with ID 1 */
+ xen,dev-domid = <1>;
+ };
--
2.7.4
On Thu, Apr 14, 2022 at 10:19:27PM +0300, Oleksandr Tyshchenko wrote:
> From: Oleksandr Tyshchenko <[email protected]>
>
> Hello all.
>
> The purpose of this RFC patch series is to add support for restricting memory access under Xen using specific
> grant table based DMA ops layer. Patch series is based on Juergen Gross’ initial work [1] which implies using
> grant references instead of raw guest physical addresses (GPA) for the virtio communications (some kind of
> the software IOMMU).
>
> The high level idea is to create new Xen’s grant table based DMA ops layer for the guest Linux whose main
> purpose is to provide a special 64-bit DMA address which is formed by using the grant reference (for a page
> to be shared with the backend) with offset and setting the highest address bit (this is for the backend to
> be able to distinguish grant ref based DMA address from normal GPA). For this to work we need the ability
> to allocate contiguous (consecutive) grant references for multi-page allocations. And the backend then needs
> to offer VIRTIO_F_ACCESS_PLATFORM and VIRTIO_F_VERSION_1 feature bits (it must support virtio-mmio modern
> transport for 64-bit addresses in the virtqueue).
I'm not enough of a xen expert to review this, and I didn't get
all patches, but I'm very happy to see that approach being
taken. VIRTIO_F_ACCESS_PLATFORM and VIRTIO_F_VERSION_1 are
exactly the way to declare not all of memory is accessible.
Thanks!
> Xen's grant mapping mechanism is the secure and safe solution to share pages between domains which proven
> to work and works for years (in the context of traditional Xen PV drivers for example). So far, the foreign
> mapping is used for the virtio backend to map and access guest memory. With the foreign mapping, the backend
> is able to map arbitrary pages from the guest memory (or even from Dom0 memory). And as the result, the malicious
> backend which runs in a non-trusted domain can take advantage of this. Instead, with the grant mapping
> the backend is only allowed to map pages which were explicitly granted by the guest before and nothing else.
> According to the discussions in various mainline threads this solution would likely be welcome because it
> perfectly fits in the security model Xen provides.
>
> What is more, the grant table based solution requires zero changes to the Xen hypervisor itself at least
> with virtio-mmio and DT (in comparison, for example, with "foreign mapping + virtio-iommu" solution which would
> require the whole new complex emulator in hypervisor in addition to new functionality/hypercall to pass IOVA
> from the virtio backend running elsewhere to the hypervisor and translate it to the GPA before mapping into
> P2M or denying the foreign mapping request if no corresponding IOVA-GPA mapping present in the IOMMU page table
> for that particular device). We only need to update toolstack to insert a new "xen,dev-domid" property to
> the virtio-mmio device node when creating a guest device-tree (this is an indicator for the guest to use grants
> and the ID of Xen domain where the corresponding backend resides, it is used as an argument to the grant mapping
> APIs). It worth mentioning that toolstack patch is based on non upstreamed yet “Virtio support for toolstack
> on Arm” series which is on review now [2].
>
> Please note the following:
> - Patch series only covers Arm and virtio-mmio (device-tree) for now. To enable the restricted memory access
> feature on Arm the following options should be set:
> CONFIG_XEN_VIRTIO = y
> CONFIG_XEN_HVM_VIRTIO_GRANT = y
> - Some callbacks in xen-virtio DMA ops layer (map_sg/unmap_sg, etc) are not implemented yet as they are not
> needed/used in the first prototype
>
> Patch series is rebased on Linux 5.18-rc2 tag and tested on Renesas Salvator-X board + H3 ES3.0 SoC (Arm64)
> with standalone userspace (non-Qemu) virtio-mmio based virtio-disk backend running in Driver domain and Linux
> guest running on existing virtio-blk driver (frontend). No issues were observed. Guest domain 'reboot/destroy'
> use-cases work properly. I have also tested other use-cases such as assigning several virtio block devices
> or a mix of virtio and Xen PV block devices to the guest.
>
> 1. Xen changes located at (last patch):
> https://github.com/otyshchenko1/xen/commits/libxl_virtio_next
> 2. Linux changes located at:
> https://github.com/otyshchenko1/linux/commits/virtio_grant5
> 3. virtio-disk changes located at:
> https://github.com/otyshchenko1/virtio-disk/commits/virtio_grant
>
> Any feedback/help would be highly appreciated.
>
> [1] https://www.youtube.com/watch?v=IrlEdaIUDPk
> [2] https://lore.kernel.org/xen-devel/[email protected]/
>
> Juergen Gross (2):
> xen/grants: support allocating consecutive grants
> virtio: add option to restrict memory access under Xen
>
> Oleksandr Tyshchenko (4):
> dt-bindings: xen: Add xen,dev-domid property description for
> xen-virtio layer
> virtio: Various updates to xen-virtio DMA ops layer
> arm/xen: Introduce xen_setup_dma_ops()
> arm/xen: Assign xen-virtio DMA ops for virtio devices in Xen guests
>
> .../devicetree/bindings/virtio/xen,dev-domid.yaml | 39 +++
> arch/arm/include/asm/xen/xen-ops.h | 1 +
> arch/arm/mm/dma-mapping.c | 5 +-
> arch/arm/xen/enlighten.c | 11 +
> arch/arm64/include/asm/xen/xen-ops.h | 1 +
> arch/arm64/mm/dma-mapping.c | 5 +-
> arch/x86/mm/init.c | 15 +
> arch/x86/mm/mem_encrypt.c | 5 -
> arch/x86/xen/Kconfig | 9 +
> drivers/xen/Kconfig | 20 ++
> drivers/xen/Makefile | 1 +
> drivers/xen/grant-table.c | 238 +++++++++++++--
> drivers/xen/xen-virtio.c | 335 +++++++++++++++++++++
> include/xen/arm/xen-ops.h | 20 ++
> include/xen/grant_table.h | 4 +
> include/xen/xen-ops.h | 13 +
> 16 files changed, 679 insertions(+), 43 deletions(-)
> create mode 100644 Documentation/devicetree/bindings/virtio/xen,dev-domid.yaml
> create mode 100644 arch/arm/include/asm/xen/xen-ops.h
> create mode 100644 arch/arm64/include/asm/xen/xen-ops.h
> create mode 100644 drivers/xen/xen-virtio.c
> create mode 100644 include/xen/arm/xen-ops.h
>
> --
> 2.7.4
On Thu, Apr 14, 2022 at 10:19:31PM +0300, Oleksandr Tyshchenko wrote:
> From: Oleksandr Tyshchenko <[email protected]>
Various updates is a big indicator that the patch should be split
further. Please do one change at a time, and fold updates to the
previous patches in the series into those patches instead of fixing
them up later.
On 16.04.22 09:05, Christoph Hellwig wrote:
Hello Christoph
> On Thu, Apr 14, 2022 at 10:19:31PM +0300, Oleksandr Tyshchenko wrote:
>> From: Oleksandr Tyshchenko <[email protected]>
> Various updates is a big indicator that the patch should be split
> further. Please do one change at a time, and fold updates to the
> previous patches in the series into those patches instead of fixing
> them up later.
Sure, next (non-RFC) version will do things properly.
--
Regards,
Oleksandr Tyshchenko
Hello Stefano, Juergen
On 18.04.22 22:11, Stefano Stabellini wrote:
> On Mon, 18 Apr 2022, Oleksandr wrote:
>> On 16.04.22 09:07, Christoph Hellwig wrote:
>>
>> Hello Christoph
>>
>>> On Fri, Apr 15, 2022 at 03:02:45PM -0700, Stefano Stabellini wrote:
>>>> This makes sense overall. Considering that the swiotlb-xen case and the
>>>> virtio case are mutually exclusive, I would write it like this:
>>> Curious question: Why can't the same grant scheme also be used for
>>> non-virtio devices? I really hate having virtio hooks in the arch
>>> dma code. Why can't Xen just say in DT/ACPI that grants can be used
>>> for a given device?
> [...]
>
>> This patch series tries to make things work with "virtio" devices in Xen
>> system without introducing any modifications to code under drivers/virtio.
>
> Actually, I think Christoph has a point.
>
> There is nothing inherently virtio specific in this patch series or in
> the "xen,dev-domid" device tree binding.
Although the main intention of this series was to enable using virtio
devices in Xen guests, I agree that nothing in new DMA ops layer
(xen-virtio.c) is virtio specific (at least at the moment). Regarding
the whole patch series I am not quite sure, as it uses
arch_has_restricted_virtio_memory_access().
> Assuming a given device is
> emulated by a Xen backend, it could be used with grants as well.
>
> For instance, we could provide an emulated e1000 NIC with a
> "xen,dev-domid" property in device tree. Linux could use grants with it
> and the backend could map the grants. It would work the same way as
> virtio-net/block/etc. Passthrough devices wouldn't have the
> "xen,dev-domid" property, so no problems.
>
> So I think we could easily generalize this work and expand it to any
> device. We just need to hook on the "xen,dev-domid" device tree
> property.
>
> I think it is just a matter of:
> - remove the "virtio,mmio" check from xen_is_virtio_device
> - rename xen_is_virtio_device to something more generic, like
> xen_is_grants_device
> - rename xen_virtio_setup_dma_ops to something more generic, like
> xen_grants_setup_dma_ops
>
> And that's pretty much it.
+ likely renaming everything in that patch series not to mention virtio
(mostly related to xen-virtio.c internals).
Stefano, thank you for clarifying Christoph's point.
Well, I am not against going this direction. Could we please make a
decision on this? @Juergen, what is your opinion?
--
Regards,
Oleksandr Tyshchenko
Hello Stefano, Juergen
On 19.04.22 09:58, Juergen Gross wrote:
> On 18.04.22 21:11, Stefano Stabellini wrote:
>> On Sun, 17 Apr 2022, Oleksandr wrote:
>>> On 16.04.22 01:02, Stefano Stabellini wrote:
>>>> On Thu, 14 Apr 2022, Oleksandr Tyshchenko wrote:
>>>>> From: Oleksandr Tyshchenko <[email protected]>
>>>>>
>>>>> In the context of current patch do the following:
>>>>> 1. Update code to support virtio-mmio devices
>>>>> 2. Introduce struct xen_virtio_data and account passed virtio devices
>>>>> (using list) as we need to store some per-device data
>>>>> 3. Add multi-page support for xen_virtio_dma_map(unmap)_page
>>>>> callbacks
>>>>> 4. Harden code against malicious backend
>>>>> 5. Change to use alloc_pages_exact() instead of __get_free_pages()
>>>>> 6. Introduce locking scheme to protect mappings (I am not 100% sure
>>>>> whether per-device lock is really needed)
>>>>> 7. Handle virtio device's DMA mask
>>>>> 8. Retrieve the ID of backend domain from DT for virtio-mmio device
>>>>> instead of hardcoding it.
>>>>>
>>>>> Signed-off-by: Oleksandr Tyshchenko <[email protected]>
>>>>> ---
>>>>> arch/arm/xen/enlighten.c | 11 +++
>>>>> drivers/xen/Kconfig | 2 +-
>>>>> drivers/xen/xen-virtio.c | 200
>>>>> ++++++++++++++++++++++++++++++++++++++++++-----
>>>>> include/xen/xen-ops.h | 5 ++
>>>>> 4 files changed, 196 insertions(+), 22 deletions(-)
>>>>>
>>>>> diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
>>>>> index ec5b082..870d92f 100644
>>>>> --- a/arch/arm/xen/enlighten.c
>>>>> +++ b/arch/arm/xen/enlighten.c
>>>>> @@ -409,6 +409,17 @@ int __init arch_xen_unpopulated_init(struct
>>>>> resource
>>>>> **res)
>>>>> }
>>>>> #endif
>>>>> +#ifdef CONFIG_ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
>>>>> +int arch_has_restricted_virtio_memory_access(void)
>>>>> +{
>>>>> + if (IS_ENABLED(CONFIG_XEN_HVM_VIRTIO_GRANT) && xen_hvm_domain())
>>>>> + return 1;
>>>> Instead of xen_hvm_domain(), you can just use xen_domain(). Also there
>>>> is no need for the #ifdef
>>>> CONFIG_ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS, given that:
>>>>
>>>> CONFIG_XEN_HVM_VIRTIO_GRANT depends on XEN_VIRTIO which selects
>>>> ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
>>>
>>>
>>> Yes, but please see my comments in commit #2 regarding
>>> CONFIG_XEN_HVM_VIRTIO_GRANT option and
>>> arch_has_restricted_virtio_memory_access() on Arm.
>>>
>>> I propose to have the following on Arm:
>>>
>>> int arch_has_restricted_virtio_memory_access(void)
>>> {
>>> return xen_has_restricted_virtio_memory_access();
>>> }
>>>
>>>
>>> where common xen.h contain a helper to be used by both Arm and x86:
>>>
>>> static inline int xen_has_restricted_virtio_memory_access(void)
>>> {
>>> if (IS_ENABLED(CONFIG_XEN_VIRTIO) && (xen_pv_domain() ||
>>> xen_hvm_domain()))
>>> return 1;
>>>
>>> return 0;
>>> }
>>>
>>>
>>> But I would be happy with what you propose as well.
>>
>> As I wrote in the previous reply, I also prefer to share the code
>> between x86 and ARM, and I think it could look like:
>>
>> int arch_has_restricted_virtio_memory_access(void)
>> {
>> return xen_has_restricted_virtio_memory_access();
>> }
>> [...]
>> static inline int xen_has_restricted_virtio_memory_access(void)
>> {
>> return (IS_ENABLED(CONFIG_XEN_VIRTIO) && xen_domain());
>> }
>>
>> But let's check with Juergen and Boris.
for the record, it is already clarified in commit #2, I will use this
variant.
>>
>>
>>
>>>>> + return 0;
>>>>> +}
>>>>> +EXPORT_SYMBOL_GPL(arch_has_restricted_virtio_memory_access);
>>>>> +#endif
>>>>> +
>>>>> static void __init xen_dt_guest_init(void)
>>>>> {
>>>>> struct device_node *xen_node;
>>>>> diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig
>>>>> index fc61f7a..56afe6a 100644
>>>>> --- a/drivers/xen/Kconfig
>>>>> +++ b/drivers/xen/Kconfig
>>>>> @@ -347,7 +347,7 @@ config XEN_VIRTIO
>>>>> config XEN_HVM_VIRTIO_GRANT
>>>>> bool "Require virtio for fully virtualized guests to use grant
>>>>> mappings"
>>>>> - depends on XEN_VIRTIO && X86_64
>>>>> + depends on XEN_VIRTIO && (X86_64 || ARM || ARM64)
>>>> you can remove the architectural dependencies
>>>
>>>
>>> According to the conversation in commit #2 we are considering just a
>>> single
>>> XEN_VIRTIO option, but it is going to has the
>>> same architectural dependencies: (X86_64 || ARM || ARM64)
>>>
>>> By removing the architectural dependencies here, we will leave also
>>> X86_32
>>> covered (neither XEN_HVM_VIRTIO_GRANT nor XEN_PV_VIRTIO covered it).
>>> I don't
>>> know whether it is ok or not.
>>>
>>> Shall I remove dependencies anyway?
>>
>> No, good point. I don't know about X86_32. This is another detail where
>> Juergen or Boris should comment.
>
> X86_32 should in theory work (it is HVM/PVH only, as PV 32-bit guests
> are no
> longer supported).
ok, thank you for confirming. I will drop architectural dependencies then.
>
>
>
> Juergen
--
Regards,
Oleksandr Tyshchenko
On 18.04.22 21:11, Stefano Stabellini wrote:
> On Sun, 17 Apr 2022, Oleksandr wrote:
>> On 16.04.22 01:02, Stefano Stabellini wrote:
>>> On Thu, 14 Apr 2022, Oleksandr Tyshchenko wrote:
>>>> From: Oleksandr Tyshchenko <[email protected]>
>>>>
>>>> In the context of current patch do the following:
>>>> 1. Update code to support virtio-mmio devices
>>>> 2. Introduce struct xen_virtio_data and account passed virtio devices
>>>> (using list) as we need to store some per-device data
>>>> 3. Add multi-page support for xen_virtio_dma_map(unmap)_page callbacks
>>>> 4. Harden code against malicious backend
>>>> 5. Change to use alloc_pages_exact() instead of __get_free_pages()
>>>> 6. Introduce locking scheme to protect mappings (I am not 100% sure
>>>> whether per-device lock is really needed)
>>>> 7. Handle virtio device's DMA mask
>>>> 8. Retrieve the ID of backend domain from DT for virtio-mmio device
>>>> instead of hardcoding it.
>>>>
>>>> Signed-off-by: Oleksandr Tyshchenko <[email protected]>
>>>> ---
>>>> arch/arm/xen/enlighten.c | 11 +++
>>>> drivers/xen/Kconfig | 2 +-
>>>> drivers/xen/xen-virtio.c | 200
>>>> ++++++++++++++++++++++++++++++++++++++++++-----
>>>> include/xen/xen-ops.h | 5 ++
>>>> 4 files changed, 196 insertions(+), 22 deletions(-)
>>>>
>>>> diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
>>>> index ec5b082..870d92f 100644
>>>> --- a/arch/arm/xen/enlighten.c
>>>> +++ b/arch/arm/xen/enlighten.c
>>>> @@ -409,6 +409,17 @@ int __init arch_xen_unpopulated_init(struct resource
>>>> **res)
>>>> }
>>>> #endif
>>>> +#ifdef CONFIG_ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
>>>> +int arch_has_restricted_virtio_memory_access(void)
>>>> +{
>>>> + if (IS_ENABLED(CONFIG_XEN_HVM_VIRTIO_GRANT) && xen_hvm_domain())
>>>> + return 1;
>>> Instead of xen_hvm_domain(), you can just use xen_domain(). Also there
>>> is no need for the #ifdef
>>> CONFIG_ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS, given that:
>>>
>>> CONFIG_XEN_HVM_VIRTIO_GRANT depends on XEN_VIRTIO which selects
>>> ARCH_HAS_RESTRICTED_VIRTIO_MEMORY_ACCESS
>>
>>
>> Yes, but please see my comments in commit #2 regarding
>> CONFIG_XEN_HVM_VIRTIO_GRANT option and
>> arch_has_restricted_virtio_memory_access() on Arm.
>>
>> I propose to have the following on Arm:
>>
>> int arch_has_restricted_virtio_memory_access(void)
>> {
>> return xen_has_restricted_virtio_memory_access();
>> }
>>
>>
>> where common xen.h contain a helper to be used by both Arm and x86:
>>
>> static inline int xen_has_restricted_virtio_memory_access(void)
>> {
>> if (IS_ENABLED(CONFIG_XEN_VIRTIO) && (xen_pv_domain() ||
>> xen_hvm_domain()))
>> return 1;
>>
>> return 0;
>> }
>>
>>
>> But I would be happy with what you propose as well.
>
> As I wrote in the previous reply, I also prefer to share the code
> between x86 and ARM, and I think it could look like:
>
> int arch_has_restricted_virtio_memory_access(void)
> {
> return xen_has_restricted_virtio_memory_access();
> }
> [...]
> static inline int xen_has_restricted_virtio_memory_access(void)
> {
> return (IS_ENABLED(CONFIG_XEN_VIRTIO) && xen_domain());
> }
>
> But let's check with Juergen and Boris.
>
>
>>>> + return 0;
>>>> +}
>>>> +EXPORT_SYMBOL_GPL(arch_has_restricted_virtio_memory_access);
>>>> +#endif
>>>> +
>>>> static void __init xen_dt_guest_init(void)
>>>> {
>>>> struct device_node *xen_node;
>>>> diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig
>>>> index fc61f7a..56afe6a 100644
>>>> --- a/drivers/xen/Kconfig
>>>> +++ b/drivers/xen/Kconfig
>>>> @@ -347,7 +347,7 @@ config XEN_VIRTIO
>>>> config XEN_HVM_VIRTIO_GRANT
>>>> bool "Require virtio for fully virtualized guests to use grant
>>>> mappings"
>>>> - depends on XEN_VIRTIO && X86_64
>>>> + depends on XEN_VIRTIO && (X86_64 || ARM || ARM64)
>>> you can remove the architectural dependencies
>>
>>
>> According to the conversation in commit #2 we are considering just a single
>> XEN_VIRTIO option, but it is going to has the
>> same architectural dependencies: (X86_64 || ARM || ARM64)
>>
>> By removing the architectural dependencies here, we will leave also X86_32
>> covered (neither XEN_HVM_VIRTIO_GRANT nor XEN_PV_VIRTIO covered it). I don't
>> know whether it is ok or not.
>>
>> Shall I remove dependencies anyway?
>
> No, good point. I don't know about X86_32. This is another detail where
> Juergen or Boris should comment.
X86_32 should in theory work (it is HVM/PVH only, as PV 32-bit guests are no
longer supported).
Juergen
Hello Stefano, Juergen
On 19.04.22 17:48, Juergen Gross wrote:
> On 19.04.22 14:17, Oleksandr wrote:
>>
>> Hello Stefano, Juergen
>>
>>
>> On 18.04.22 22:11, Stefano Stabellini wrote:
>>> On Mon, 18 Apr 2022, Oleksandr wrote:
>>>> On 16.04.22 09:07, Christoph Hellwig wrote:
>>>>
>>>> Hello Christoph
>>>>
>>>>> On Fri, Apr 15, 2022 at 03:02:45PM -0700, Stefano Stabellini wrote:
>>>>>> This makes sense overall. Considering that the swiotlb-xen case
>>>>>> and the
>>>>>> virtio case are mutually exclusive, I would write it like this:
>>>>> Curious question: Why can't the same grant scheme also be used for
>>>>> non-virtio devices? I really hate having virtio hooks in the arch
>>>>> dma code. Why can't Xen just say in DT/ACPI that grants can be used
>>>>> for a given device?
>>> [...]
>>>
>>>> This patch series tries to make things work with "virtio" devices
>>>> in Xen
>>>> system without introducing any modifications to code under
>>>> drivers/virtio.
>>>
>>> Actually, I think Christoph has a point.
>>>
>>> There is nothing inherently virtio specific in this patch series or in
>>> the "xen,dev-domid" device tree binding.
>>
>>
>> Although the main intention of this series was to enable using virtio
>> devices in Xen guests, I agree that nothing in new DMA ops layer
>> (xen-virtio.c) is virtio specific (at least at the moment). Regarding
>> the whole patch series I am not quite sure, as it uses
>> arch_has_restricted_virtio_memory_access(). >
>>> Assuming a given device is
>>> emulated by a Xen backend, it could be used with grants as well.
>>>
>>> For instance, we could provide an emulated e1000 NIC with a
>>> "xen,dev-domid" property in device tree. Linux could use grants with it
>>> and the backend could map the grants. It would work the same way as
>>> virtio-net/block/etc. Passthrough devices wouldn't have the
>>> "xen,dev-domid" property, so no problems.
>>>
>>> So I think we could easily generalize this work and expand it to any
>>> device. We just need to hook on the "xen,dev-domid" device tree
>>> property.
>>>
>>> I think it is just a matter of:
>>> - remove the "virtio,mmio" check from xen_is_virtio_device
>>> - rename xen_is_virtio_device to something more generic, like
>>> xen_is_grants_device
>
> xen_is_grants_dma_device, please. Normal Xen PV devices are covered by
> grants, too, and I'd like to avoid the confusion arising from this.
yes, this definitely makes sense as we need to distinguish
>
>
>>> - rename xen_virtio_setup_dma_ops to something more generic, like
>>> xen_grants_setup_dma_ops
>>>
>>> And that's pretty much it.
>>
>> + likely renaming everything in that patch series not to mention
>> virtio (mostly related to xen-virtio.c internals).
>>
>>
>> Stefano, thank you for clarifying Christoph's point.
>>
>> Well, I am not against going this direction. Could we please make a
>> decision on this? @Juergen, what is your opinion?
>
> Yes, why not.
ok, thank you for confirming.
>
>
> Maybe rename xen-virtio.c to grant-dma.c?
Personally I don't mind.
>
> I'd keep the XEN_VIRTIO related config option, as this will be the
> normal use
> case. grant-dma.c should be covered by a new hidden config option
> XEN_GRANT_DMA
> selected by XEN_VIRTIO.
I got it, ok
>
>
> CONFIG_XEN_VIRTIO should still guard
> xen_has_restricted_virtio_memory_access().
ok
So a few questions to clarify:
1. What is the best place to keep "xen,dev-domid" binding's description
now? I think that proposed in current series place
(Documentation/devicetree/bindings/virtio/) is not good fit now.
2. I assume the logic in the current patch will remain the same, I mean
we will still assign Xen grant DMA ops from xen_setup_dma_ops() here?
>
>
>
> Juergen
--
Regards,
Oleksandr Tyshchenko
On Wed, 20 Apr 2022, Oleksandr wrote:
> On 20.04.22 03:23, Stefano Stabellini wrote:
> > On Tue, 19 Apr 2022, Oleksandr wrote:
> > > On 19.04.22 17:48, Juergen Gross wrote:
> > > > On 19.04.22 14:17, Oleksandr wrote:
> > > > > Hello Stefano, Juergen
> > > > >
> > > > >
> > > > > On 18.04.22 22:11, Stefano Stabellini wrote:
> > > > > > On Mon, 18 Apr 2022, Oleksandr wrote:
> > > > > > > On 16.04.22 09:07, Christoph Hellwig wrote:
> > > > > > >
> > > > > > > Hello Christoph
> > > > > > >
> > > > > > > > On Fri, Apr 15, 2022 at 03:02:45PM -0700, Stefano Stabellini
> > > > > > > > wrote:
> > > > > > > > > This makes sense overall. Considering that the swiotlb-xen
> > > > > > > > > case
> > > > > > > > > and the
> > > > > > > > > virtio case are mutually exclusive, I would write it like
> > > > > > > > > this:
> > > > > > > > Curious question: Why can't the same grant scheme also be used
> > > > > > > > for
> > > > > > > > non-virtio devices? I really hate having virtio hooks in the
> > > > > > > > arch
> > > > > > > > dma code. Why can't Xen just say in DT/ACPI that grants can be
> > > > > > > > used
> > > > > > > > for a given device?
> > > > > > [...]
> > > > > >
> > > > > > > This patch series tries to make things work with "virtio" devices
> > > > > > > in
> > > > > > > Xen
> > > > > > > system without introducing any modifications to code under
> > > > > > > drivers/virtio.
> > > > > > Actually, I think Christoph has a point.
> > > > > >
> > > > > > There is nothing inherently virtio specific in this patch series or
> > > > > > in
> > > > > > the "xen,dev-domid" device tree binding.
> > > > >
> > > > > Although the main intention of this series was to enable using virtio
> > > > > devices in Xen guests, I agree that nothing in new DMA ops layer
> > > > > (xen-virtio.c) is virtio specific (at least at the moment). Regarding
> > > > > the
> > > > > whole patch series I am not quite sure, as it uses
> > > > > arch_has_restricted_virtio_memory_access(). >
> > > > > > Assuming a given device is
> > > > > > emulated by a Xen backend, it could be used with grants as well.
> > > > > >
> > > > > > For instance, we could provide an emulated e1000 NIC with a
> > > > > > "xen,dev-domid" property in device tree. Linux could use grants with
> > > > > > it
> > > > > > and the backend could map the grants. It would work the same way as
> > > > > > virtio-net/block/etc. Passthrough devices wouldn't have the
> > > > > > "xen,dev-domid" property, so no problems.
> > > > > >
> > > > > > So I think we could easily generalize this work and expand it to any
> > > > > > device. We just need to hook on the "xen,dev-domid" device tree
> > > > > > property.
> > > > > >
> > > > > > I think it is just a matter of:
> > > > > > - remove the "virtio,mmio" check from xen_is_virtio_device
> > > > > > - rename xen_is_virtio_device to something more generic, like
> > > > > > xen_is_grants_device
> > > > xen_is_grants_dma_device, please. Normal Xen PV devices are covered by
> > > > grants, too, and I'd like to avoid the confusion arising from this.
> > >
> > > yes, this definitely makes sense as we need to distinguish
> > >
> > >
> > > >
> > > > > > - rename xen_virtio_setup_dma_ops to something more generic, like
> > > > > > xen_grants_setup_dma_ops
> > > > > >
> > > > > > And that's pretty much it.
> > > > > + likely renaming everything in that patch series not to mention
> > > > > virtio
> > > > > (mostly related to xen-virtio.c internals).
> > > > >
> > > > >
> > > > > Stefano, thank you for clarifying Christoph's point.
> > > > >
> > > > > Well, I am not against going this direction. Could we please make a
> > > > > decision on this? @Juergen, what is your opinion?
> > > > Yes, why not.
> > >
> > > ok, thank you for confirming.
> > >
> > >
> > > >
> > > > Maybe rename xen-virtio.c to grant-dma.c?
> > >
> > > Personally I don't mind.
> > >
> > >
> > > > I'd keep the XEN_VIRTIO related config option, as this will be the
> > > > normal
> > > > use
> > > > case. grant-dma.c should be covered by a new hidden config option
> > > > XEN_GRANT_DMA
> > > > selected by XEN_VIRTIO.
> > >
> > > I got it, ok
> > >
> > >
> > > >
> > > > CONFIG_XEN_VIRTIO should still guard
> > > > xen_has_restricted_virtio_memory_access().
> > >
> > > ok
> > >
> > >
> > > So a few questions to clarify:
> > >
> > > 1. What is the best place to keep "xen,dev-domid" binding's description
> > > now? I
> > > think that proposed in current series place
> > > (Documentation/devicetree/bindings/virtio/) is not good fit now.
> > I would probably add it to the existing
> > Documentation/devicetree/bindings/arm/xen.txt.
> >
> >
> > > 2. I assume the logic in the current patch will remain the same, I mean we
> > > will still assign Xen grant DMA ops from xen_setup_dma_ops() here?
> > Yes I think so
>
>
> Stefano, thank you for clarifying!
>
>
> Regarding new naming scheme...
>
> As there is an existing Kconfig option XEN_GRANT_DMA_ALLOC used for different
> purpose, we need to clarify naming scheme here a bit to avoid possible
> confusion.
>
> For example, I am happy with proposed by Juergen ...
>
> ... Kconfig option: XEN_GRANT_DMA_OPS
>
> and
>
> ... file: grant-dma-ops.c
I think that's fine by me
On 19.04.22 14:17, Oleksandr wrote:
>
> Hello Stefano, Juergen
>
>
> On 18.04.22 22:11, Stefano Stabellini wrote:
>> On Mon, 18 Apr 2022, Oleksandr wrote:
>>> On 16.04.22 09:07, Christoph Hellwig wrote:
>>>
>>> Hello Christoph
>>>
>>>> On Fri, Apr 15, 2022 at 03:02:45PM -0700, Stefano Stabellini wrote:
>>>>> This makes sense overall. Considering that the swiotlb-xen case and the
>>>>> virtio case are mutually exclusive, I would write it like this:
>>>> Curious question: Why can't the same grant scheme also be used for
>>>> non-virtio devices? I really hate having virtio hooks in the arch
>>>> dma code. Why can't Xen just say in DT/ACPI that grants can be used
>>>> for a given device?
>> [...]
>>
>>> This patch series tries to make things work with "virtio" devices in Xen
>>> system without introducing any modifications to code under drivers/virtio.
>>
>> Actually, I think Christoph has a point.
>>
>> There is nothing inherently virtio specific in this patch series or in
>> the "xen,dev-domid" device tree binding.
>
>
> Although the main intention of this series was to enable using virtio devices in
> Xen guests, I agree that nothing in new DMA ops layer (xen-virtio.c) is virtio
> specific (at least at the moment). Regarding the whole patch series I am not
> quite sure, as it uses arch_has_restricted_virtio_memory_access(). >
>> Assuming a given device is
>> emulated by a Xen backend, it could be used with grants as well.
>>
>> For instance, we could provide an emulated e1000 NIC with a
>> "xen,dev-domid" property in device tree. Linux could use grants with it
>> and the backend could map the grants. It would work the same way as
>> virtio-net/block/etc. Passthrough devices wouldn't have the
>> "xen,dev-domid" property, so no problems.
>>
>> So I think we could easily generalize this work and expand it to any
>> device. We just need to hook on the "xen,dev-domid" device tree
>> property.
>>
>> I think it is just a matter of:
>> - remove the "virtio,mmio" check from xen_is_virtio_device
>> - rename xen_is_virtio_device to something more generic, like
>> xen_is_grants_device
xen_is_grants_dma_device, please. Normal Xen PV devices are covered by
grants, too, and I'd like to avoid the confusion arising from this.
>> - rename xen_virtio_setup_dma_ops to something more generic, like
>> xen_grants_setup_dma_ops
>>
>> And that's pretty much it.
>
> + likely renaming everything in that patch series not to mention virtio (mostly
> related to xen-virtio.c internals).
>
>
> Stefano, thank you for clarifying Christoph's point.
>
> Well, I am not against going this direction. Could we please make a decision on
> this? @Juergen, what is your opinion?
Yes, why not.
Maybe rename xen-virtio.c to grant-dma.c?
I'd keep the XEN_VIRTIO related config option, as this will be the normal use
case. grant-dma.c should be covered by a new hidden config option XEN_GRANT_DMA
selected by XEN_VIRTIO.
CONFIG_XEN_VIRTIO should still guard xen_has_restricted_virtio_memory_access().
Juergen
On Tue, 19 Apr 2022, Oleksandr wrote:
> On 19.04.22 17:48, Juergen Gross wrote:
> > On 19.04.22 14:17, Oleksandr wrote:
> > >
> > > Hello Stefano, Juergen
> > >
> > >
> > > On 18.04.22 22:11, Stefano Stabellini wrote:
> > > > On Mon, 18 Apr 2022, Oleksandr wrote:
> > > > > On 16.04.22 09:07, Christoph Hellwig wrote:
> > > > >
> > > > > Hello Christoph
> > > > >
> > > > > > On Fri, Apr 15, 2022 at 03:02:45PM -0700, Stefano Stabellini wrote:
> > > > > > > This makes sense overall. Considering that the swiotlb-xen case
> > > > > > > and the
> > > > > > > virtio case are mutually exclusive, I would write it like this:
> > > > > > Curious question: Why can't the same grant scheme also be used for
> > > > > > non-virtio devices? I really hate having virtio hooks in the arch
> > > > > > dma code. Why can't Xen just say in DT/ACPI that grants can be used
> > > > > > for a given device?
> > > > [...]
> > > >
> > > > > This patch series tries to make things work with "virtio" devices in
> > > > > Xen
> > > > > system without introducing any modifications to code under
> > > > > drivers/virtio.
> > > >
> > > > Actually, I think Christoph has a point.
> > > >
> > > > There is nothing inherently virtio specific in this patch series or in
> > > > the "xen,dev-domid" device tree binding.
> > >
> > >
> > > Although the main intention of this series was to enable using virtio
> > > devices in Xen guests, I agree that nothing in new DMA ops layer
> > > (xen-virtio.c) is virtio specific (at least at the moment). Regarding the
> > > whole patch series I am not quite sure, as it uses
> > > arch_has_restricted_virtio_memory_access(). >
> > > > Assuming a given device is
> > > > emulated by a Xen backend, it could be used with grants as well.
> > > >
> > > > For instance, we could provide an emulated e1000 NIC with a
> > > > "xen,dev-domid" property in device tree. Linux could use grants with it
> > > > and the backend could map the grants. It would work the same way as
> > > > virtio-net/block/etc. Passthrough devices wouldn't have the
> > > > "xen,dev-domid" property, so no problems.
> > > >
> > > > So I think we could easily generalize this work and expand it to any
> > > > device. We just need to hook on the "xen,dev-domid" device tree
> > > > property.
> > > >
> > > > I think it is just a matter of:
> > > > - remove the "virtio,mmio" check from xen_is_virtio_device
> > > > - rename xen_is_virtio_device to something more generic, like
> > > > xen_is_grants_device
> >
> > xen_is_grants_dma_device, please. Normal Xen PV devices are covered by
> > grants, too, and I'd like to avoid the confusion arising from this.
>
>
> yes, this definitely makes sense as we need to distinguish
>
>
> >
> >
> > > > - rename xen_virtio_setup_dma_ops to something more generic, like
> > > > xen_grants_setup_dma_ops
> > > >
> > > > And that's pretty much it.
> > >
> > > + likely renaming everything in that patch series not to mention virtio
> > > (mostly related to xen-virtio.c internals).
> > >
> > >
> > > Stefano, thank you for clarifying Christoph's point.
> > >
> > > Well, I am not against going this direction. Could we please make a
> > > decision on this? @Juergen, what is your opinion?
> >
> > Yes, why not.
>
>
> ok, thank you for confirming.
>
>
> >
> >
> > Maybe rename xen-virtio.c to grant-dma.c?
>
>
> Personally I don't mind.
>
>
> >
> > I'd keep the XEN_VIRTIO related config option, as this will be the normal
> > use
> > case. grant-dma.c should be covered by a new hidden config option
> > XEN_GRANT_DMA
> > selected by XEN_VIRTIO.
>
>
> I got it, ok
>
>
> >
> >
> > CONFIG_XEN_VIRTIO should still guard
> > xen_has_restricted_virtio_memory_access().
>
>
> ok
>
>
> So a few questions to clarify:
>
> 1. What is the best place to keep "xen,dev-domid" binding's description now? I
> think that proposed in current series place
> (Documentation/devicetree/bindings/virtio/) is not good fit now.
I would probably add it to the existing
Documentation/devicetree/bindings/arm/xen.txt.
> 2. I assume the logic in the current patch will remain the same, I mean we
> will still assign Xen grant DMA ops from xen_setup_dma_ops() here?
Yes I think so
Hello Stefano, Juergen
On 20.04.22 03:23, Stefano Stabellini wrote:
> On Tue, 19 Apr 2022, Oleksandr wrote:
>> On 19.04.22 17:48, Juergen Gross wrote:
>>> On 19.04.22 14:17, Oleksandr wrote:
>>>> Hello Stefano, Juergen
>>>>
>>>>
>>>> On 18.04.22 22:11, Stefano Stabellini wrote:
>>>>> On Mon, 18 Apr 2022, Oleksandr wrote:
>>>>>> On 16.04.22 09:07, Christoph Hellwig wrote:
>>>>>>
>>>>>> Hello Christoph
>>>>>>
>>>>>>> On Fri, Apr 15, 2022 at 03:02:45PM -0700, Stefano Stabellini wrote:
>>>>>>>> This makes sense overall. Considering that the swiotlb-xen case
>>>>>>>> and the
>>>>>>>> virtio case are mutually exclusive, I would write it like this:
>>>>>>> Curious question: Why can't the same grant scheme also be used for
>>>>>>> non-virtio devices? I really hate having virtio hooks in the arch
>>>>>>> dma code. Why can't Xen just say in DT/ACPI that grants can be used
>>>>>>> for a given device?
>>>>> [...]
>>>>>
>>>>>> This patch series tries to make things work with "virtio" devices in
>>>>>> Xen
>>>>>> system without introducing any modifications to code under
>>>>>> drivers/virtio.
>>>>> Actually, I think Christoph has a point.
>>>>>
>>>>> There is nothing inherently virtio specific in this patch series or in
>>>>> the "xen,dev-domid" device tree binding.
>>>>
>>>> Although the main intention of this series was to enable using virtio
>>>> devices in Xen guests, I agree that nothing in new DMA ops layer
>>>> (xen-virtio.c) is virtio specific (at least at the moment). Regarding the
>>>> whole patch series I am not quite sure, as it uses
>>>> arch_has_restricted_virtio_memory_access(). >
>>>>> Assuming a given device is
>>>>> emulated by a Xen backend, it could be used with grants as well.
>>>>>
>>>>> For instance, we could provide an emulated e1000 NIC with a
>>>>> "xen,dev-domid" property in device tree. Linux could use grants with it
>>>>> and the backend could map the grants. It would work the same way as
>>>>> virtio-net/block/etc. Passthrough devices wouldn't have the
>>>>> "xen,dev-domid" property, so no problems.
>>>>>
>>>>> So I think we could easily generalize this work and expand it to any
>>>>> device. We just need to hook on the "xen,dev-domid" device tree
>>>>> property.
>>>>>
>>>>> I think it is just a matter of:
>>>>> - remove the "virtio,mmio" check from xen_is_virtio_device
>>>>> - rename xen_is_virtio_device to something more generic, like
>>>>> xen_is_grants_device
>>> xen_is_grants_dma_device, please. Normal Xen PV devices are covered by
>>> grants, too, and I'd like to avoid the confusion arising from this.
>>
>> yes, this definitely makes sense as we need to distinguish
>>
>>
>>>
>>>>> - rename xen_virtio_setup_dma_ops to something more generic, like
>>>>> xen_grants_setup_dma_ops
>>>>>
>>>>> And that's pretty much it.
>>>> + likely renaming everything in that patch series not to mention virtio
>>>> (mostly related to xen-virtio.c internals).
>>>>
>>>>
>>>> Stefano, thank you for clarifying Christoph's point.
>>>>
>>>> Well, I am not against going this direction. Could we please make a
>>>> decision on this? @Juergen, what is your opinion?
>>> Yes, why not.
>>
>> ok, thank you for confirming.
>>
>>
>>>
>>> Maybe rename xen-virtio.c to grant-dma.c?
>>
>> Personally I don't mind.
>>
>>
>>> I'd keep the XEN_VIRTIO related config option, as this will be the normal
>>> use
>>> case. grant-dma.c should be covered by a new hidden config option
>>> XEN_GRANT_DMA
>>> selected by XEN_VIRTIO.
>>
>> I got it, ok
>>
>>
>>>
>>> CONFIG_XEN_VIRTIO should still guard
>>> xen_has_restricted_virtio_memory_access().
>>
>> ok
>>
>>
>> So a few questions to clarify:
>>
>> 1. What is the best place to keep "xen,dev-domid" binding's description now? I
>> think that proposed in current series place
>> (Documentation/devicetree/bindings/virtio/) is not good fit now.
> I would probably add it to the existing
> Documentation/devicetree/bindings/arm/xen.txt.
>
>
>> 2. I assume the logic in the current patch will remain the same, I mean we
>> will still assign Xen grant DMA ops from xen_setup_dma_ops() here?
> Yes I think so
Stefano, thank you for clarifying!
Regarding new naming scheme...
As there is an existing Kconfig option XEN_GRANT_DMA_ALLOC used for
different purpose, we need to clarify naming scheme here a bit to avoid
possible confusion.
For example, I am happy with proposed by Juergen ...
... Kconfig option: XEN_GRANT_DMA_OPS
and
... file: grant-dma-ops.c
--
Regards,
Oleksandr Tyshchenko