CXL: Taint user access to DOE mailbox config space
v3 cover CXL: Taint user access to DOE mailbox config space
From: Ira Weiny <[email protected]>
Changes from V2
Incorporate feedback from Greg and Jonathan
Changes from V1
Incorporate feedback from Dan and Greg.
PCI config space access from user space has traditionally been unrestricted with writes being an understood risk for device operation.
Unfortunately, device breakage or odd behavior from config writes lacks indicators that can leave driver writers confused when evaluating failures. This is especially true with the new PCIe Data Object Exchange (DOE) mailbox protocol where backdoor shenanigans from user space through things such as vendor defined protocols may affect device operation without complete breakage.
Even though access should not be restricted it would be nice for driver writers to be able to flag critical parts of the config space such that interference from user space can be detected.
Introduce pci_request_config_region_exclusive() and use it in the CXL driver for DOE config space.
Ira Weiny (2):
PCI: Allow drivers to request exclusive config regions cxl/doe: Request exclusive DOE access
drivers/cxl/pci.c | 5 +++++ drivers/pci/pci-sysfs.c | 7 +++++++ drivers/pci/probe.c | 6 ++++++ include/linux/ioport.h | 2 ++ include/linux/pci.h | 17 +++++++++++++++++ include/uapi/linux/pci_regs.h | 1 + kernel/resource.c | 13 ++++++++----- 7 files changed, 46 insertions(+), 5 deletions(-)
base-commit: f76349cf41451c5c42a99f18a9163377e4b364ff
Signature · 1 line
2.37.2
PATCH v3 1/2 PCI: Allow drivers to request exclusive config regions
From: Ira Weiny <[email protected]>
PCI config space access from user space has traditionally been unrestricted with writes being an understood risk for device operation.
Unfortunately, device breakage or odd behavior from config writes lacks indicators that can leave driver writers confused when evaluating failures. This is especially true with the new PCIe Data Object Exchange (DOE) mailbox protocol where backdoor shenanigans from user space through things such as vendor defined protocols may affect device operation without complete breakage.
A prior proposal restricted read and writes completely.[1] Greg and Bjorn pointed out that proposal is flawed for a couple of reasons. First, lspci should always be allowed and should not interfere with any device operation. Second, setpci is a valuable tool that is sometimes necessary and it should not be completely restricted.[2] Finally methods exist for full lock of device access if required.
Even though access should not be restricted it would be nice for driver writers to be able to flag critical parts of the config space such that interference from user space can be detected.
Introduce pci_request_config_region_exclusive() to mark exclusive config regions. Such regions trigger a warning and kernel taint if accessed via user space.
Create pci_warn_once() to restrict the user from spamming the log.
[1] https://lore.kernel.org/all/161663543465.1867664.5674061943008380442.stgit@dwillia2-desk3.amr.corp.intel.com/
[2] https://lore.kernel.org/all/[email protected]/
---
Changes from V2:
Greg: s/config_resource/driver_exclusive_resource/
Jonathan: don't reformat the pci_*() messages
Changes from V1:
Greg and Dan:
Create and use pci_warn_once() to keep the user from spamming
Dan:
Clarify the warn message
Changes from[1]:
Change name to pci_request_config_region_exclusive() Don't flag reads at all.
Allow writes with a warn and taint of the kernel. Update commit message
Forward port to latest tree.
--- drivers/pci/pci-sysfs.c | 7 +++++++ drivers/pci/probe.c | 6 ++++++ include/linux/ioport.h | 2 ++ include/linux/pci.h | 17 +++++++++++++++++ kernel/resource.c | 13 ++++++++----- 5 files changed, 40 insertions(+), 5 deletions(-)
drivers/pci/pci-sysfs.c +7 -0
@@ -755,6 +755,13 @@ static ssize_t pci_write_config(struct file *filp, struct kobject *kobj,if (ret)return ret; +if (resource_is_exclusive (&dev->driver_exclusive_resource, off, + count)) { +pci_warn_once (dev,"%s: Unexpected write to kernel-exclusive config offset %llx" , + current->comm, off); +add_taint (TAINT_USER ,LOCKDEP_STILL_OK ); + } +if (off > dev->cfg_size)return 0 ;if (off + count > dev->cfg_size) {
drivers/pci/probe.c +6 -0
@@ -2306,6 +2306,12 @@ struct pci_dev *pci_alloc_dev(struct pci_bus *bus)INIT_LIST_HEAD (&dev->bus_list); dev->dev.type = &pci_dev_type; dev->bus =pci_bus_get (bus); + dev->driver_exclusive_resource = (struct resource ) { + .name ="PCI Exclusive" , + .start =0 , + .end =-1 , + }; +#ifdef CONFIG_PCI_MSI raw_spin_lock_init (&dev->msi_lock);#endif
include/linux/ioport.h +2 -0
@@ -312,6 +312,8 @@ extern void __devm_release_region(struct device *dev, struct resource *parent, resource_size_t start, resource_size_t n);extern int iomem_map_sanity_check (resource_size_t addr,unsigned long size);extern bool iomem_is_exclusive (u64 addr); +extern bool resource_is_exclusive (struct resource *resource,u64 addr, +resource_size_t size); extern int walk_system_ram_range(unsigned long start_pfn, unsigned long nr_pages,
include/linux/pci.h +17 -0
@@ -409,6 +409,7 @@ struct pci_dev {*/ unsigned int irq;struct resource resource[DEVICE_COUNT_RESOURCE ];/* I/O and memory regions + expansion ROMs */ +struct resource driver_exclusive_resource;/* driver exclusive resource ranges */ bool match_driver;/* Skip attaching driver */
@@ -1406,6 +1407,21 @@ int pci_request_selected_regions(struct pci_dev *, int, const char *);int pci_request_selected_regions_exclusive (struct pci_dev *,int ,const char *);void pci_release_selected_regions (struct pci_dev *,int ); +static inline __must_check struct resource * +pci_request_config_region_exclusive (struct pci_dev *pdev,unsigned int offset, +unsigned int len,const char *name) +{ +return __request_region (&pdev->driver_exclusive_resource, offset, len, + name,IORESOURCE_EXCLUSIVE ); +} + +static inline void pci_release_config_region (struct pci_dev *pdev, +unsigned int offset, +unsigned int len) +{ +__release_region (&pdev->driver_exclusive_resource, offset, len); +} +/* drivers/pci/bus.c */ void pci_add_resource (struct list_head *resources,struct resource *res); void pci_add_resource_offset(struct list_head *resources, struct resource *res,
@@ -2481,6 +2497,7 @@ void pci_uevent_ers(struct pci_dev *pdev, enum pci_ers_result err_type); #define pci_crit(pdev, fmt, arg...) dev_crit(&(pdev)->dev, fmt, ##arg) #define pci_err(pdev, fmt, arg...) dev_err(&(pdev)->dev, fmt, ##arg) #define pci_warn(pdev, fmt, arg...) dev_warn(&(pdev)->dev, fmt, ##arg) +#define pci_warn_once(pdev, fmt, arg...) dev_warn_once(&(pdev)->dev, fmt, ##arg) #define pci_notice(pdev, fmt, arg...) dev_notice(&(pdev)->dev, fmt, ##arg) #define pci_info(pdev, fmt, arg...) dev_info(&(pdev)->dev, fmt, ##arg) #define pci_dbg(pdev, fmt, arg...) dev_dbg(&(pdev)->dev, fmt, ##arg)
kernel/resource.c +8 -5
@@ -1707,18 +1707,15 @@ static int strict_iomem_checks;* * Returns true if exclusive to the kernel, otherwise returns false. */ -bool iomem_is_exclusive (u64 addr) +bool resource_is_exclusive (struct resource *root,u64 addr,resource_size_t size) {const unsigned int exclusive_system_ram =IORESOURCE_SYSTEM_RAM |IORESOURCE_EXCLUSIVE ;bool skip_children = false, err = false; -int size =PAGE_SIZE ;struct resource *p; - addr = addr &PAGE_MASK ; -read_lock (&resource_lock); -for_each_resource (&iomem_resource, p, skip_children) { +for_each_resource (root, p, skip_children) {if (p->start >= addr + size)break ;if (p->end < addr) {
@@ -1757,6 +1754,12 @@ bool iomem_is_exclusive(u64 addr)return err; } +bool iomem_is_exclusive (u64 addr) +{ +return resource_is_exclusive (&iomem_resource, addr &PAGE_MASK , +PAGE_SIZE ); +} +struct resource_entry *resource_list_create_entry (struct resource *res,size_t extra_size) {
Signature · 1 line
2.37.2
Re: [PATCH V3 1/2] PCI: Allow drivers to request exclusive config regions
Quoted 29 lines · On Mon, Sep 26, 2022 at 02:57:10PM -0700, [email protected] wrote:
From: Ira Weiny <[email protected]>
PCI config space access from user space has traditionally been unrestricted with writes being an understood risk for device operation.
Unfortunately, device breakage or odd behavior from config writes lacks indicators that can leave driver writers confused when evaluating failures. This is especially true with the new PCIe Data Object Exchange (DOE) mailbox protocol where backdoor shenanigans from user space through things such as vendor defined protocols may affect device operation without complete breakage.
A prior proposal restricted read and writes completely.[1] Greg and Bjorn pointed out that proposal is flawed for a couple of reasons. First, lspci should always be allowed and should not interfere with any device operation. Second, setpci is a valuable tool that is sometimes necessary and it should not be completely restricted.[2] Finally methods exist for full lock of device access if required.
Even though access should not be restricted it would be nice for driver writers to be able to flag critical parts of the config space such that interference from user space can be detected.
Introduce
pci_request_config_region_exclusive()to mark exclusive config regions. Such regions trigger a warning and kernel taint if accessed via user space.Create
pci_warn_once()to restrict the user from spamming the log.
Quoted 8 lines
[1] https://lore.kernel.org/all/161663543465.1867664.5674061943008380442.stgit@dwillia2-desk3.amr.corp.intel.com/
[2] https://lore.kernel.org/all/[email protected]/Cc Bjorn Helgaas <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Reviewed-by Jonathan Cameron <[email protected]>Suggested-by Dan Williams <[email protected]>Signed-off-by Ira Weiny <[email protected]>
Re: [PATCH V3 1/2] PCI: Allow drivers to request exclusive config regions
In reply to [email protected]Quoted 29 lines · On Mon, Sep 26, 2022 at 02:57:10PM -0700, [email protected] wrote:
From: Ira Weiny <[email protected]>
PCI config space access from user space has traditionally been unrestricted with writes being an understood risk for device operation.
Unfortunately, device breakage or odd behavior from config writes lacks indicators that can leave driver writers confused when evaluating failures. This is especially true with the new PCIe Data Object Exchange (DOE) mailbox protocol where backdoor shenanigans from user space through things such as vendor defined protocols may affect device operation without complete breakage.
A prior proposal restricted read and writes completely.[1] Greg and Bjorn pointed out that proposal is flawed for a couple of reasons. First, lspci should always be allowed and should not interfere with any device operation. Second, setpci is a valuable tool that is sometimes necessary and it should not be completely restricted.[2] Finally methods exist for full lock of device access if required.
Even though access should not be restricted it would be nice for driver writers to be able to flag critical parts of the config space such that interference from user space can be detected.
Introduce
pci_request_config_region_exclusive()to mark exclusive config regions. Such regions trigger a warning and kernel taint if accessed via user space.Create
pci_warn_once()to restrict the user from spamming the log.
Quoted 8 lines
[1] https://lore.kernel.org/all/161663543465.1867664.5674061943008380442.stgit@dwillia2-desk3.amr.corp.intel.com/
[2] https://lore.kernel.org/all/[email protected]/Cc Bjorn Helgaas <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Reviewed-by Jonathan Cameron <[email protected]>Suggested-by Dan Williams <[email protected]>Signed-off-by Ira Weiny <[email protected]>
Quoted 157 lines
---
Changes from V2:
Greg: s/config_resource/driver_exclusive_resource/
Jonathan: don't reformat the pci_*() messagesChanges from V1:
Greg and Dan:
Create and usepci_warn_once()to keep the user from spamming
Dan:
Clarify the warn messageChanges from[1]:
Change name topci_request_config_region_exclusive()Don't flag reads at all.
Allow writes with a warn and taint of the kernel. Update commit message
Forward port to latest tree.--- drivers/pci/pci-sysfs.c | 7 +++++++ drivers/pci/probe.c | 6 ++++++ include/linux/ioport.h | 2 ++ include/linux/pci.h | 17 +++++++++++++++++ kernel/resource.c | 13 ++++++++----- 5 files changed, 40 insertions(+), 5 deletions(-)drivers/pci/pci-sysfs.c +7
-0@@ -755,6 +755,13 @@ static ssize_t pci_write_config(struct file *filp, struct kobject *kobj,if (ret)return ret; +if (resource_is_exclusive (&dev->driver_exclusive_resource, off, + count)) { +pci_warn_once (dev,"%s: Unexpected write to kernel-exclusive config offset %llx" , + current->comm, off); +add_taint (TAINT_USER ,LOCKDEP_STILL_OK ); + } +if (off > dev->cfg_size)return 0 ;if (off + count > dev->cfg_size) {drivers/pci/probe.c +6
-0@@ -2306,6 +2306,12 @@ struct pci_dev *pci_alloc_dev(struct pci_bus *bus)INIT_LIST_HEAD (&dev->bus_list); dev->dev.type = &pci_dev_type; dev->bus =pci_bus_get (bus); + dev->driver_exclusive_resource = (struct resource ) { + .name ="PCI Exclusive" , + .start =0 , + .end =-1 , + }; +#ifdef CONFIG_PCI_MSI raw_spin_lock_init (&dev->msi_lock);#endif include/linux/ioport.h +2
-0@@ -312,6 +312,8 @@ extern void __devm_release_region(struct device *dev, struct resource *parent, resource_size_t start, resource_size_t n);extern int iomem_map_sanity_check (resource_size_t addr,unsigned long size);extern bool iomem_is_exclusive (u64 addr); +extern bool resource_is_exclusive (struct resource *resource,u64 addr, +resource_size_t size); extern int walk_system_ram_range(unsigned long start_pfn, unsigned long nr_pages,include/linux/pci.h +17
-0@@ -409,6 +409,7 @@ struct pci_dev {*/ unsigned int irq;struct resource resource[DEVICE_COUNT_RESOURCE ];/* I/O and memory regions + expansion ROMs */ +struct resource driver_exclusive_resource;/* driver exclusive resource ranges */ bool match_driver;/* Skip attaching driver */ @@ -1406,6 +1407,21 @@ int pci_request_selected_regions(struct pci_dev *, int, const char *);int pci_request_selected_regions_exclusive (struct pci_dev *,int ,const char *);void pci_release_selected_regions (struct pci_dev *,int ); +static inline __must_check struct resource * +pci_request_config_region_exclusive (struct pci_dev *pdev,unsigned int offset, +unsigned int len,const char *name) +{ +return __request_region (&pdev->driver_exclusive_resource, offset, len, + name,IORESOURCE_EXCLUSIVE ); +} + +static inline void pci_release_config_region (struct pci_dev *pdev, +unsigned int offset, +unsigned int len) +{ +__release_region (&pdev->driver_exclusive_resource, offset, len); +} +/* drivers/pci/bus.c */ void pci_add_resource (struct list_head *resources,struct resource *res); void pci_add_resource_offset(struct list_head *resources, struct resource *res,@@ -2481,6 +2497,7 @@ void pci_uevent_ers(struct pci_dev *pdev, enum pci_ers_result err_type); #define pci_crit(pdev, fmt, arg...) dev_crit(&(pdev)->dev, fmt, ##arg) #define pci_err(pdev, fmt, arg...) dev_err(&(pdev)->dev, fmt, ##arg) #define pci_warn(pdev, fmt, arg...) dev_warn(&(pdev)->dev, fmt, ##arg) +#define pci_warn_once(pdev, fmt, arg...) dev_warn_once(&(pdev)->dev, fmt, ##arg) #define pci_notice(pdev, fmt, arg...) dev_notice(&(pdev)->dev, fmt, ##arg) #define pci_info(pdev, fmt, arg...) dev_info(&(pdev)->dev, fmt, ##arg) #define pci_dbg(pdev, fmt, arg...) dev_dbg(&(pdev)->dev, fmt, ##arg)kernel/resource.c +8
-5@@ -1707,18 +1707,15 @@ static int strict_iomem_checks;* * Returns true if exclusive to the kernel, otherwise returns false. */ -bool iomem_is_exclusive (u64 addr) +bool resource_is_exclusive (struct resource *root,u64 addr,resource_size_t size) {const unsigned int exclusive_system_ram =IORESOURCE_SYSTEM_RAM |IORESOURCE_EXCLUSIVE ;bool skip_children = false, err = false; -int size =PAGE_SIZE ;struct resource *p; - addr = addr &PAGE_MASK ; -read_lock (&resource_lock); -for_each_resource (&iomem_resource, p, skip_children) { +for_each_resource (root, p, skip_children) {if (p->start >= addr + size)break ;if (p->end < addr) {@@ -1757,6 +1754,12 @@ bool iomem_is_exclusive(u64 addr)return err; } +bool iomem_is_exclusive (u64 addr) +{ +return resource_is_exclusive (&iomem_resource, addr &PAGE_MASK , +PAGE_SIZE ); +} +struct resource_entry *resource_list_create_entry (struct resource *res,size_t extra_size) {Signature · 1 line
2.37.2
PATCH v3 2/2 cxl/doe: Request exclusive DOE access
In reply to [email protected]From: Ira Weiny <[email protected]>
The PCIE Data Object Exchange (DOE) mailbox is a protocol run over configuration cycles. It assumes one initiator at a time. While the kernel has control of the mailbox user space writes could interfere with the kernel access.
Mark DOE mailbox config space exclusive when iterated by the CXL driver.
---
Changes from V2:
Jonathan:
s/PCI_DOE_CAP_SIZE/PCI_DOE_CAP_SIZEOF Set PCI_DOE_CAP_SIZEOF directly
--- drivers/cxl/pci.c | 5 +++++ include/uapi/linux/pci_regs.h | 1 + 2 files changed, 6 insertions(+)
drivers/cxl/pci.c +5 -0
@@ -418,6 +418,11 @@ static void devm_cxl_pci_create_doe(struct cxl_dev_state *cxlds)continue ; } +if (!pci_request_config_region_exclusive (pdev, off, +PCI_DOE_CAP_SIZEOF , +dev_name (dev))) +pci_err (pdev,"Failed to exclude DOE registers\n" ); +if (xa_insert (&cxlds->doe_mbs, off, doe_mb,GFP_KERNEL )) {dev_err (dev,"xa_insert failed to insert MB @ %x\n" , off);
include/uapi/linux/pci_regs.h +1 -0
@@ -1119,6 +1119,7 @@#define PCI_DOE_STATUS_DATA_OBJECT_READY 0x80000000/* Data Object Ready */ #define PCI_DOE_WRITE 0x10/* DOE Write Data Mailbox Register */ #define PCI_DOE_READ 0x14/* DOE Read Data Mailbox Register */ +#define PCI_DOE_CAP_SIZEOF 0x18/* Size of DOE register block */ /* DOE Data Object - note not actually registers */ #define PCI_DOE_DATA_OBJECT_HEADER_1_VID 0x0000ffff
Signature · 1 line
2.37.2
Re: [PATCH V3 2/2] cxl/doe: Request exclusive DOE access
Quoted 10 lines · On Mon, 26 Sep 2022 14:57:11 -0700 [email protected] wrote:
From: Ira Weiny <[email protected]>
The PCIE Data Object Exchange (DOE) mailbox is a protocol run over configuration cycles. It assumes one initiator at a time. While the kernel has control of the mailbox user space writes could interfere with the kernel access.
Mark DOE mailbox config space exclusive when iterated by the CXL driver.
Signed-off-by Ira Weiny <[email protected]>
I wonder a bit on whether the failure should be fatal given that something very odd would be required for it to fail.
I'm not that bothered though.
Quoted 39 lines
---
Changes from V2:
Jonathan:
s/PCI_DOE_CAP_SIZE/PCI_DOE_CAP_SIZEOF SetPCI_DOE_CAP_SIZEOFdirectly--- drivers/cxl/pci.c | 5 +++++ include/uapi/linux/pci_regs.h | 1 + 2 files changed, 6 insertions(+)drivers/cxl/pci.c +5
-0@@ -418,6 +418,11 @@ static void devm_cxl_pci_create_doe(struct cxl_dev_state *cxlds)continue ; } +if (!pci_request_config_region_exclusive (pdev, off, +PCI_DOE_CAP_SIZEOF , +dev_name (dev))) +pci_err (pdev,"Failed to exclude DOE registers\n" ); +if (xa_insert (&cxlds->doe_mbs, off, doe_mb,GFP_KERNEL )) {dev_err (dev,"xa_insert failed to insert MB @ %x\n" , off);include/uapi/linux/pci_regs.h +1
-0@@ -1119,6 +1119,7 @@#define PCI_DOE_STATUS_DATA_OBJECT_READY 0x80000000/* Data Object Ready */ #define PCI_DOE_WRITE 0x10/* DOE Write Data Mailbox Register */ #define PCI_DOE_READ 0x14/* DOE Read Data Mailbox Register */ +#define PCI_DOE_CAP_SIZEOF 0x18/* Size of DOE register block */ /* DOE Data Object - note not actually registers */ #define PCI_DOE_DATA_OBJECT_HEADER_1_VID 0x0000ffff
Quoted 38 lines · ira.weiny@ wrote:
From: Ira Weiny <[email protected]>
Changes from V2
Incorporate feedback from Greg and JonathanChanges from V1
Incorporate feedback from Dan and Greg.PCI config space access from user space has traditionally been unrestricted with writes being an understood risk for device operation.
Unfortunately, device breakage or odd behavior from config writes lacks indicators that can leave driver writers confused when evaluating failures. This is especially true with the new PCIe Data Object Exchange (DOE) mailbox protocol where backdoor shenanigans from user space through things such as vendor defined protocols may affect device operation without complete breakage.
Even though access should not be restricted it would be nice for driver writers to be able to flag critical parts of the config space such that interference from user space can be detected.
Introduce
pci_request_config_region_exclusive()and use it in the CXL driver for DOE config space.Ira Weiny (2):
PCI: Allow drivers to request exclusive config regions cxl/doe: Request exclusive DOE accessdrivers/cxl/pci.c | 5 +++++ drivers/pci/pci-sysfs.c | 7 +++++++ drivers/pci/probe.c | 6 ++++++ include/linux/ioport.h | 2 ++ include/linux/pci.h | 17 +++++++++++++++++ include/uapi/linux/pci_regs.h | 1 + kernel/resource.c | 13 ++++++++----- 7 files changed, 46 insertions(+), 5 deletions(-)
Quoted 1 line
base-commit:
f76349cf41451c5c42a99f18a9163377e4b364ff
Applied for v6.2. I am not a huge fan of the driver_exclusive_resource naming since it leaves out the configuration address space aspect. However, any other naming I can think of confuses it with the traditional PCI device resources which do more than just host exclusions.