Fix PF/VF dependency issue
v7 cover Fix PF/VF dependency issue
From: Kuppuswamy Sathyanarayanan <[email protected]>
Current implementation of ATS, PASID, PRI does not handle VF dependencies correctly. Following patches addresses this issue.
Changes since v6:
* Removed locking interfaces used in v6.
* Made necessary changes to PRI/PASID enable/disable calls to allow register changes only for PF.
Changes since v5:
* Created new patches for PRI/PASID capability caching.
* Removed individual locks (pri_lock, pasid_lock) and added common resource lock to synchronize PRI/PASID updates between PF/VF.
* Addressed comments from Bjorn Helgaas.
Changes since v4:
* Defined empty functions for pci_pri_init() and pci_pasid_init() for cases where CONFIG_PCI_PRI and CONFIG_PCI_PASID are not enabled.
Changes since v3:
* Fixed critical path (lock context) in pci_restore_*_state functions.
Changes since v2:
* Added locking mechanism to synchronize accessing PF registers in VF.
* Removed spec compliance checks in patches.
* Addressed comments from Bjorn Helgaas.
Changes since v1:
* Added more details about the patches in commit log.
* Removed bulk spec check patch.
* Addressed comments from Bjorn Helgaas.
Kuppuswamy Sathyanarayanan (7):
PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues PCI/ATS: Cache PRI capability check result
PCI/ATS: Cache PASID capability check result
PCI/ATS: Add PRI support for PCIe VF devices
PCI/ATS: Add PASID support for PCIe VF devices PCI/ATS: Disable PF/VF ATS service independently
PCI: Skip Enhanced Allocation (EA) initialization for VF device
drivers/pci/ats.c | 202 ++++++++++++++++++++++++++++------------ drivers/pci/pci.c | 7 ++ include/linux/pci-ats.h | 12 ++- include/linux/pci.h | 3 +- 4 files changed, 159 insertions(+), 65 deletions(-)
Signature · 1 line
2.21.0
PATCH v7 1/7 PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since pci_prg_resp_pasid_required() function has dependency on both PASID and PRI, define it only if both CONFIG_PCI_PRI and CONFIG_PCI_PASID config options are enabled.
interface.")
--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)
drivers/pci/ats.c +6 -4
@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status.
@@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {
@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/**
include/linux/pci-ats.h +9 -3
@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */
@@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/
Signature · 1 line
2.21.0
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix e5567f5f6762. That commit added pci_prg_resp_pasid_required(), but with no dependency on CONFIG_PCI_PRI or CONFIG_PCI_PASID.
This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
Quoted 77 lines
Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)drivers/pci/ats.c +6
-4@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status. @@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/** include/linux/pci-ats.h +9
-3@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */ @@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/ Signature · 1 line
2.21.0
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Hi Bjorn,
Thanks for looking into this patch set.
Quoted 15 lines · Bjorn Helgaas
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix
e5567f5f6762. That commit addedpci_prg_resp_pasid_required(), but with no dependency onCONFIG_PCI_PRIorCONFIG_PCI_PASID.
Quoted 6 lines
This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
In your v8 submission you did not merge this patch. You did not use pri_cap or pasid_cap cached values. Instead you have re-read the value from register. Is this intentional?
Since this function will be called for every VF device we might loose some performance benefit.
Quoted 78 lines
Quoted 78 lines
Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)drivers/pci/ats.c +6
-4@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status. @@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/** include/linux/pci-ats.h +9
-3@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */ @@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/ Signature · 1 line
2.21.0
Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Quoted 24 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Thanks for looking into this patch set.
Quoted 19 lines · Bjorn Helgaas
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix
e5567f5f6762. That commit addedpci_prg_resp_pasid_required(), but with no dependency onCONFIG_PCI_PRIorCONFIG_PCI_PASID.This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
Quoted 8 lines
Quoted 2 lines
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
In your v8 submission you did not merge this patch. You did not use
pri_caporpasid_capcached values. Instead you have re-read the value from register. Is this intentional?Since this function will be called for every VF device we might loose some performance benefit.
This particular patch doesn't do any caching. IIRC it fiddles with ifdefs to solve a problem that would be introduced by a future patch. I don't remember the exact details, but I think the series I merged doesn't have that problem. If it does, let me know the details and we can fix it.
I did include the caching patches for both PRI and PASID capabilities, but they're only performance optimizations so I moved them to the end so the functional fixes would be smaller and earlier in the series.
Quoted 75 lines
Quoted 71 lines
Quoted 71 lines
Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)drivers/pci/ats.c +6
-4@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status. @@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/** include/linux/pci-ats.h +9
-3@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */ @@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/ Signature · 1 line
2.21.0Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Quoted 31 lines · Bjorn Helgaas
Quoted 30 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Thanks for looking into this patch set.
Quoted 15 lines · Bjorn Helgaas
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix
e5567f5f6762. That commit addedpci_prg_resp_pasid_required(), but with no dependency onCONFIG_PCI_PRIorCONFIG_PCI_PASID.Quoted 6 lines
This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
In your v8 submission you did not merge this patch. You did not use
pri_caporpasid_capcached values. Instead you have re-read the value from register. Is this intentional?
Quoted 8 lines
Quoted 2 lines
Since this function will be called for every VF device we might loose some performance benefit.
This particular patch doesn't do any caching. IIRC it fiddles with ifdefs to solve a problem that would be introduced by a future patch. I don't remember the exact details, but I think the series I merged doesn't have that problem. If it does, let me know the details and we can fix it.
This patch by itself does not do any caching. But your caching patch missed modifying this function to use cached values. Please check the current implementation of this function. It still reads PCI_EXT_CAP_ID_PRI register instead of using cached value. Please let me know your comments.
int pci_prg_resp_pasid_required(struct pci_dev *pdev)
{
u16 status;
int pri;
if (pdev->is_virtfn)
pdev = pci_physfn(pdev);
pri = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
if (!pri)
return 0;
pci_read_config_word(pdev, pri + PCI_PRI_STATUS, &status);
if (status & PCI_PRI_STATUS_PASID)
return 1;
return 0;
}EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);
If caching is applied to this function then we need this #ifdef dependency correction patch.
Quoted 80 lines
I did include the caching patches for both PRI and PASID capabilities, but they're only performance optimizations so I moved them to the end so the functional fixes would be smaller and earlier in the series.
Quoted 75 lines
Quoted 71 lines
Quoted 71 lines
Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)drivers/pci/ats.c +6
-4@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status. @@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/** include/linux/pci-ats.h +9
-3@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */ @@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/ Signature · 1 line
2.21.0Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Quoted 65 lines · Kuppuswamy Sathyanarayanan
Quoted 31 lines · Bjorn Helgaas
Quoted 30 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Thanks for looking into this patch set.
Quoted 15 lines · Bjorn Helgaas
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix
e5567f5f6762. That commit addedpci_prg_resp_pasid_required(), but with no dependency onCONFIG_PCI_PRIorCONFIG_PCI_PASID.Quoted 6 lines
This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
In your v8 submission you did not merge this patch. You did not use
pri_caporpasid_capcached values. Instead you have re-read the value from register. Is this intentional?Quoted 8 lines
Quoted 2 lines
Since this function will be called for every VF device we might loose some performance benefit.
This particular patch doesn't do any caching. IIRC it fiddles with ifdefs to solve a problem that would be introduced by a future patch. I don't remember the exact details, but I think the series I merged doesn't have that problem. If it does, let me know the details and we can fix it.
This patch by itself does not do any caching. But your caching patch missed modifying this function to use cached values. Please check the current implementation of this function. It still reads
PCI_EXT_CAP_ID_PRIregister instead of using cached value. Please let me know your comments.int
pci_prg_resp_pasid_required(struct pci_dev *pdev){ u16 status; int pri; if (pdev->is_virtfn) pdev = pci_physfn(pdev); pri = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI); if (!pri) return 0; pci_read_config_word(pdev, pri + PCI_PRI_STATUS, &status); if (status & PCI_PRI_STATUS_PASID) return 1; return 0; }
Quoted 4 lines
EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);If caching is applied to this function then we need this #ifdef dependency correction patch.
IIRC this #ifdef patch wasn't connected to the actual *need* for the #ifdef, so it was very difficult to review. I thought this function would be infrequently used and it wasn't worth trying to sort out the #ifdef muddle to do the caching. But it does seem sort of pointless to chase the capability list again here, so maybe it *is* worth optimizing.
The PRG Response PASID Required bit is read-only, so I wonder if it would be simpler if we just read PCI_PRI_STATUS once and save the bit in the struct pci_dev? We could do that in pci_enable_pri(), or if we might need the value before that's called, we could add a pci_pri_init() and do it there.
Quoted 83 lines
Quoted 79 lines
I did include the caching patches for both PRI and PASID capabilities, but they're only performance optimizations so I moved them to the end so the functional fixes would be smaller and earlier in the series.
Quoted 75 lines
Quoted 71 lines
Quoted 71 lines
Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)drivers/pci/ats.c +6
-4@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status. @@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/** include/linux/pci-ats.h +9
-3@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */ @@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/ Signature · 1 line
2.21.0Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developerSignature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Quoted 76 lines · Bjorn Helgaas
Quoted 64 lines · Kuppuswamy Sathyanarayanan
Quoted 31 lines · Bjorn Helgaas
Quoted 30 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Thanks for looking into this patch set.
Quoted 15 lines · Bjorn Helgaas
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix
e5567f5f6762. That commit addedpci_prg_resp_pasid_required(), but with no dependency onCONFIG_PCI_PRIorCONFIG_PCI_PASID.Quoted 6 lines
This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
In your v8 submission you did not merge this patch. You did not use
pri_caporpasid_capcached values. Instead you have re-read the value from register. Is this intentional?Quoted 7 lines
Quoted 2 lines
Since this function will be called for every VF device we might loose some performance benefit.
This particular patch doesn't do any caching. IIRC it fiddles with ifdefs to solve a problem that would be introduced by a future patch. I don't remember the exact details, but I think the series I merged doesn't have that problem. If it does, let me know the details and we can fix it.
This patch by itself does not do any caching. But your caching patch missed modifying this function to use cached values. Please check the current implementation of this function. It still reads
PCI_EXT_CAP_ID_PRIregister instead of using cached value. Please let me know your comments.int
pci_prg_resp_pasid_required(struct pci_dev *pdev){ u16 status; int pri; if (pdev->is_virtfn) pdev = pci_physfn(pdev); pri = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI); if (!pri) return 0; pci_read_config_word(pdev, pri + PCI_PRI_STATUS, &status); if (status & PCI_PRI_STATUS_PASID) return 1; return 0; }Quoted 4 lines
EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);If caching is applied to this function then we need this #ifdef dependency correction patch.
IIRC this #ifdef patch wasn't connected to the actual *need* for the #ifdef, so it was very difficult to review. I thought this function would be infrequently used and it wasn't worth trying to sort out the #ifdef muddle to do the caching. But it does seem sort of pointless to chase the capability list again here, so maybe it *is* worth optimizing.
Quoted 5 lines
The PRG Response PASID Required bit is read-only, so I wonder if it would be simpler if we just read
PCI_PRI_STATUSonce and save the bit in the structpci_dev? We could do that inpci_enable_pri(), or if we might need the value before that's called, we could add apci_pri_init()and do it there.
Yes, caching PASID Required bit in pci_pri_init() function would provide performance benefits. But another thing to consider is, since this bit is same for both PF/VF, is it worth to add this bit it to struct pci_dev?or struct pci_sriov is the more appropriate place?
Quoted 83 lines
Quoted 83 lines
Quoted 80 lines
I did include the caching patches for both PRI and PASID capabilities, but they're only performance optimizations so I moved them to the end so the functional fixes would be smaller and earlier in the series.
Quoted 75 lines
Quoted 71 lines
Quoted 71 lines
Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>--- drivers/pci/ats.c | 10 ++++++---- include/linux/pci-ats.h | 12 +++++++++--- 2 files changed, 15 insertions(+), 7 deletions(-)drivers/pci/ats.c +6
-4@@ -395,6 +395,8 @@ int pci_pasid_features(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_pasid_features); +#ifdef CONFIG_PCI_PRI +/** * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit * status. @@ -402,10 +404,8 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);* * Returns 1 if PASID is required in PRG Response Message, 0 otherwise. * -* Even though the PRG response PASID status is read from PRI Status -* Register, since this API will mainly be used by PASID users, this -* function is defined within #ifdef CONFIG_PCI_PASID instead of -* CONFIG_PCI_PRI. +* Since this API has dependency on both PRI and PASID, protect it +* with both CONFIG_PCI_PRI and CONFIG_PCI_PASID. */ int pci_prg_resp_pasid_required (struct pci_dev *pdev) {@@ -425,6 +425,8 @@ int pci_prg_resp_pasid_required(struct pci_dev *pdev) }EXPORT_SYMBOL_GPL (pci_prg_resp_pasid_required); +#endif +#define PASID_NUMBER_SHIFT 8#define PASID_NUMBER_MASK (0x1f << PASID_NUMBER_SHIFT)/** include/linux/pci-ats.h +9
-3@@ -40,7 +40,6 @@ void pci_disable_pasid(struct pci_dev *pdev);void pci_restore_pasid_state (struct pci_dev *pdev);int pci_pasid_features (struct pci_dev *pdev);int pci_max_pasids (struct pci_dev *pdev); -int pci_prg_resp_pasid_required (struct pci_dev *pdev);#else /* CONFIG_PCI_PASID */ @@ -67,11 +66,18 @@ static inline int pci_max_pasids(struct pci_dev *pdev)return -EINVAL ; } +#endif /* CONFIG_PCI_PASID */ + +#if defined(CONFIG_PCI_PRI ) && defined(CONFIG_PCI_PASID ) + +int pci_prg_resp_pasid_required (struct pci_dev *pdev); + +#else /* CONFIG_PCI_PASID && CONFIG_PCI_PRI */ +static inline int pci_prg_resp_pasid_required (struct pci_dev *pdev) {return 0 ; } -#endif /* CONFIG_PCI_PASID */ - +#endif #endif /* LINUX_PCI_ATS_H*/ Signature · 1 line
2.21.0Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developerSignature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 1/7] PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
Quoted 84 lines · Kuppuswamy Sathyanarayanan
Quoted 76 lines · Bjorn Helgaas
Quoted 64 lines · Kuppuswamy Sathyanarayanan
Quoted 31 lines · Bjorn Helgaas
Quoted 30 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Thanks for looking into this patch set.
Quoted 15 lines · Bjorn Helgaas
Quoted 8 lines · On Wed, Aug 28, 2019 at 03:14:01PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Since
pci_prg_resp_pasid_required()function has dependency on both PASID and PRI, define it only if bothCONFIG_PCI_PRIandCONFIG_PCI_PASIDconfig options are enabled.Fixes e5567f5f6762 ("PCI/ATS: Add pci_prg_resp_pasid_required()interface.")
[Don't split tags, including "Fixes:" across lines]
This definitely doesn't fix
e5567f5f6762. That commit addedpci_prg_resp_pasid_required(), but with no dependency onCONFIG_PCI_PRIorCONFIG_PCI_PASID.Quoted 6 lines
This patch is only required when a subsequent patch is applied. It should be squashed into the commit that requires it so it's obvious why it's needed.
I've been poking at this series, and I'll post a v8 soon with this and other fixes.
In your v8 submission you did not merge this patch. You did not use
pri_caporpasid_capcached values. Instead you have re-read the value from register. Is this intentional?Quoted 7 lines
Quoted 2 lines
Since this function will be called for every VF device we might loose some performance benefit.
This particular patch doesn't do any caching. IIRC it fiddles with ifdefs to solve a problem that would be introduced by a future patch. I don't remember the exact details, but I think the series I merged doesn't have that problem. If it does, let me know the details and we can fix it.
This patch by itself does not do any caching. But your caching patch missed modifying this function to use cached values. Please check the current implementation of this function. It still reads
PCI_EXT_CAP_ID_PRIregister instead of using cached value. Please let me know your comments.int
pci_prg_resp_pasid_required(struct pci_dev *pdev){ u16 status; int pri; if (pdev->is_virtfn) pdev = pci_physfn(pdev); pri = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI); if (!pri) return 0; pci_read_config_word(pdev, pri + PCI_PRI_STATUS, &status); if (status & PCI_PRI_STATUS_PASID) return 1; return 0; }Quoted 4 lines
EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);If caching is applied to this function then we need this #ifdef dependency correction patch.
IIRC this #ifdef patch wasn't connected to the actual *need* for the #ifdef, so it was very difficult to review. I thought this function would be infrequently used and it wasn't worth trying to sort out the #ifdef muddle to do the caching. But it does seem sort of pointless to chase the capability list again here, so maybe it *is* worth optimizing.
Quoted 5 lines
The PRG Response PASID Required bit is read-only, so I wonder if it would be simpler if we just read
PCI_PRI_STATUSonce and save the bit in the structpci_dev? We could do that inpci_enable_pri(), or if we might need the value before that's called, we could add apci_pri_init()and do it there.
Quoted 5 lines
Yes, caching PASID Required bit in
pci_pri_init()function would provide performance benefits. But another thing to consider is, since this bit is same for both PF/VF, is it worth to add this bit it to struct pci_dev?or structpci_sriovis the more appropriate place?
IIUC, the PRI capability is not specific to SR-IOV, so I don't think it would make sense to cache PRG Response PASID Required in pci_sriov.
PFs may implement PRI; VFs do not (PCIe r5.0, sec 10.5.2), so I think the bit should be cached in the pci_dev, and if we want to know the value for a VF, we should read it from the PF's pci_dev.
Bjorn
PATCH v7 2/7 PCI/ATS: Cache PRI capability check result
In reply to [email protected]From: Kuppuswamy Sathyanarayanan <[email protected]>
Currently, PRI capability checks are repeated across all PRI API's. Instead, cache the capability check result in pci_pri_init() and use it in other PRI API's.
--- drivers/pci/ats.c | 56 +++++++++++++++++++++++++-------------------- include/linux/pci.h | 1 + 2 files changed, 32 insertions(+), 25 deletions(-)
drivers/pci/ats.c +31 -25
@@ -16,6 +16,19 @@#include "pci.h" +static void pci_pri_init (struct pci_dev *pdev) +{ +#ifdef CONFIG_PCI_PRI +int pos; + + pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI ); +if (!pos) +return ; + + pdev->pri_cap = pos; +#endif +} +void pci_ats_init (struct pci_dev *dev) {int pos;
@@ -28,6 +41,8 @@ void pci_ats_init(struct pci_dev *dev)return ; dev->ats_cap = pos; + +pci_pri_init (dev); }/**
@@ -180,26 +195,25 @@ int pci_enable_pri(struct pci_dev *pdev, u32 reqs) {u16 control, status;u32 max_requests; -int pos;if (WARN_ON (pdev->pri_enabled))return -EBUSY ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI ); -if (!pos) +if (!pdev->pri_cap)return -EINVAL ; -pci_read_config_word (pdev, pos +PCI_PRI_STATUS , &status); +pci_read_config_word (pdev, pdev->pri_cap +PCI_PRI_STATUS , &status);if (!(status &PCI_PRI_STATUS_STOPPED ))return -EBUSY ; -pci_read_config_dword (pdev, pos +PCI_PRI_MAX_REQ , &max_requests); +pci_read_config_dword (pdev, pdev->pri_cap +PCI_PRI_MAX_REQ , + &max_requests); reqs =min (max_requests, reqs); pdev->pri_reqs_alloc = reqs; -pci_write_config_dword (pdev, pos +PCI_PRI_ALLOC_REQ , reqs); +pci_write_config_dword (pdev, pdev->pri_cap +PCI_PRI_ALLOC_REQ , reqs); control =PCI_PRI_CTRL_ENABLE ; -pci_write_config_word (pdev, pos +PCI_PRI_CTRL , control); +pci_write_config_word (pdev, pdev->pri_cap +PCI_PRI_CTRL , control); pdev->pri_enabled =1 ;
@@ -216,18 +230,16 @@ EXPORT_SYMBOL_GPL(pci_enable_pri);void pci_disable_pri (struct pci_dev *pdev) {u16 control; -int pos;if (WARN_ON (!pdev->pri_enabled))return ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI ); -if (!pos) +if (!pdev->pri_cap)return ; -pci_read_config_word (pdev, pos +PCI_PRI_CTRL , &control); +pci_read_config_word (pdev, pdev->pri_cap +PCI_PRI_CTRL , &control); control &= ~PCI_PRI_CTRL_ENABLE ; -pci_write_config_word (pdev, pos +PCI_PRI_CTRL , control); +pci_write_config_word (pdev, pdev->pri_cap +PCI_PRI_CTRL , control); pdev->pri_enabled =0 ; }
@@ -241,17 +253,15 @@ void pci_restore_pri_state(struct pci_dev *pdev) {u16 control =PCI_PRI_CTRL_ENABLE ;u32 reqs = pdev->pri_reqs_alloc; -int pos;if (!pdev->pri_enabled)return ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI ); -if (!pos) +if (!pdev->pri_cap)return ; -pci_write_config_dword (pdev, pos +PCI_PRI_ALLOC_REQ , reqs); -pci_write_config_word (pdev, pos +PCI_PRI_CTRL , control); +pci_write_config_dword (pdev, pdev->pri_cap +PCI_PRI_ALLOC_REQ , reqs); +pci_write_config_word (pdev, pdev->pri_cap +PCI_PRI_CTRL , control); }EXPORT_SYMBOL_GPL (pci_restore_pri_state);
@@ -265,17 +275,15 @@ EXPORT_SYMBOL_GPL(pci_restore_pri_state);int pci_reset_pri (struct pci_dev *pdev) {u16 control; -int pos;if (WARN_ON (pdev->pri_enabled))return -EBUSY ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI ); -if (!pos) +if (!pdev->pri_cap)return -EINVAL ; control =PCI_PRI_CTRL_RESET ; -pci_write_config_word (pdev, pos +PCI_PRI_CTRL , control); +pci_write_config_word (pdev, pdev->pri_cap +PCI_PRI_CTRL , control);return 0 ; }
@@ -410,13 +418,11 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);int pci_prg_resp_pasid_required (struct pci_dev *pdev) {u16 status; -int pos; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI ); -if (!pos) +if (!pdev->pri_cap)return 0 ; -pci_read_config_word (pdev, pos +PCI_PRI_STATUS , &status); +pci_read_config_word (pdev, pdev->pri_cap +PCI_PRI_STATUS , &status);if (status &PCI_PRI_STATUS_PASID )return 1 ;
include/linux/pci.h +1 -0
@@ -455,6 +455,7 @@ struct pci_dev {atomic_t ats_ref_cnt;/* Number of VFs with ATS enabled */ #endif #ifdef CONFIG_PCI_PRI +u16 pri_cap;/* PRI Capability offset */ u32 pri_reqs_alloc;/* Number of PRI requests allocated */ #endif #ifdef CONFIG_PCI_PASID
Signature · 1 line
2.21.0
PATCH v7 3/7 PCI/ATS: Cache PASID capability check result
In reply to [email protected]From: Kuppuswamy Sathyanarayanan <[email protected]>
Currently, PASID capability checks are repeated across all PASID API's. Instead, cache the capability check result in pci_pasid_init() and use it in other PASID API's.
--- drivers/pci/ats.c | 50 ++++++++++++++++++++++++++------------------- include/linux/pci.h | 1 + 2 files changed, 30 insertions(+), 21 deletions(-)
drivers/pci/ats.c +29 -21
@@ -29,6 +29,19 @@ static void pci_pri_init(struct pci_dev *pdev)#endif } +static void pci_pasid_init (struct pci_dev *pdev) +{ +#ifdef CONFIG_PCI_PASID +int pos; + + pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID ); +if (!pos) +return ; + + pdev->pasid_cap = pos; +#endif +} +void pci_ats_init (struct pci_dev *dev) {int pos;
@@ -43,6 +56,8 @@ void pci_ats_init(struct pci_dev *dev) dev->ats_cap = pos;pci_pri_init (dev); + +pci_pasid_init (dev); }/**
@@ -303,7 +318,6 @@ EXPORT_SYMBOL_GPL(pci_reset_pri);int pci_enable_pasid (struct pci_dev *pdev,int features) {u16 control, supported; -int pos;if (WARN_ON (pdev->pasid_enabled))return -EBUSY ;
@@ -311,11 +325,11 @@ int pci_enable_pasid(struct pci_dev *pdev, int features)if (!pdev->eetlp_prefix_path)return -EINVAL ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID ); -if (!pos) +if (!pdev->pasid_cap)return -EINVAL ; -pci_read_config_word (pdev, pos +PCI_PASID_CAP , &supported); +pci_read_config_word (pdev, pdev->pasid_cap +PCI_PASID_CAP , + &supported); supported &=PCI_PASID_CAP_EXEC |PCI_PASID_CAP_PRIV ;/* User wants to enable anything unsupported? */
@@ -325,7 +339,7 @@ int pci_enable_pasid(struct pci_dev *pdev, int features) control =PCI_PASID_CTRL_ENABLE | features; pdev->pasid_features = features; -pci_write_config_word (pdev, pos +PCI_PASID_CTRL , control); +pci_write_config_word (pdev, pdev->pasid_cap +PCI_PASID_CTRL , control); pdev->pasid_enabled =1 ;
@@ -340,16 +354,14 @@ EXPORT_SYMBOL_GPL(pci_enable_pasid);void pci_disable_pasid (struct pci_dev *pdev) {u16 control =0 ; -int pos;if (WARN_ON (!pdev->pasid_enabled))return ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID ); -if (!pos) +if (!pdev->pasid_cap)return ; -pci_write_config_word (pdev, pos +PCI_PASID_CTRL , control); +pci_write_config_word (pdev, pdev->pasid_cap +PCI_PASID_CTRL , control); pdev->pasid_enabled =0 ; }
@@ -362,17 +374,15 @@ EXPORT_SYMBOL_GPL(pci_disable_pasid);void pci_restore_pasid_state (struct pci_dev *pdev) {u16 control; -int pos;if (!pdev->pasid_enabled)return ; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID ); -if (!pos) +if (!pdev->pasid_cap)return ; control =PCI_PASID_CTRL_ENABLE | pdev->pasid_features; -pci_write_config_word (pdev, pos +PCI_PASID_CTRL , control); +pci_write_config_word (pdev, pdev->pasid_cap +PCI_PASID_CTRL , control); }EXPORT_SYMBOL_GPL (pci_restore_pasid_state);
@@ -389,13 +399,12 @@ EXPORT_SYMBOL_GPL(pci_restore_pasid_state);int pci_pasid_features (struct pci_dev *pdev) {u16 supported; -int pos; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID ); -if (!pos) +if (!pdev->pasid_cap)return -EINVAL ; -pci_read_config_word (pdev, pos +PCI_PASID_CAP , &supported); +pci_read_config_word (pdev, pdev->pasid_cap +PCI_PASID_CAP , + &supported); supported &=PCI_PASID_CAP_EXEC |PCI_PASID_CAP_PRIV ;
@@ -445,13 +454,12 @@ EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);int pci_max_pasids (struct pci_dev *pdev) {u16 supported; -int pos; - pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID ); -if (!pos) +if (!pdev->pasid_cap)return -EINVAL ; -pci_read_config_word (pdev, pos +PCI_PASID_CAP , &supported); +pci_read_config_word (pdev, pdev->pasid_cap +PCI_PASID_CAP , + &supported); supported = (supported &PASID_NUMBER_MASK ) >>PASID_NUMBER_SHIFT ;
include/linux/pci.h +1 -0
@@ -459,6 +459,7 @@ struct pci_dev {u32 pri_reqs_alloc;/* Number of PRI requests allocated */ #endif #ifdef CONFIG_PCI_PASID +u16 pasid_cap;/* PASID Capability offset */ u16 pasid_features;#endif #ifdef CONFIG_PCI_P2PDMA
Signature · 1 line
2.21.0
PATCH v7 4/7 PCI/ATS: Add PRI support for PCIe VF devices
In reply to [email protected]From: Kuppuswamy Sathyanarayanan <[email protected]>
When IOMMU tries to enable Page Request Interface (PRI) for VF device in iommu_enable_dev_iotlb(), it always fails because PRI support for PCIe VF device is currently broken. Current implementation expects the given PCIe device (PF & VF) to implement PRI capability before enabling the PRI support. But this assumption is incorrect. As per PCIe spec r4.0, sec 9.3.7.11, all VFs associated with PF can only use the PRI of the PF and not implement it. Hence we need to create exception for handling the PRI support for PCIe VF device.
--- drivers/pci/ats.c | 42 ++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 40 insertions(+), 2 deletions(-)
drivers/pci/ats.c +40 -2
@@ -21,6 +21,15 @@ static void pci_pri_init(struct pci_dev *pdev)#ifdef CONFIG_PCI_PRI int pos; +/* +* As per PCIe r4.0, sec 9.3.7.11, only PF is permitted to +* implement PRI and all associated VFs can only use it. +* Since PF already initialized the PRI parameters there is +* no need to proceed further. +*/ +if (pdev->is_virtfn) +return ; + pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PRI );if (!pos)return ;
@@ -210,6 +219,20 @@ int pci_enable_pri(struct pci_dev *pdev, u32 reqs) {u16 control, status;u32 max_requests; +struct pci_dev *pf =pci_physfn (pdev); + +/* +* IOMMU is the only user of this function and as per +* current usage, PF PRI enable always happens before +* VF and hence we don't need to do anything special +* for VF. So just return success if PRI is enabled in PF. +*/ +if (pdev->is_virtfn) { +if (pf->pri_enabled) +return 0 ; +else +return -EINVAL ; + }if (WARN_ON (pdev->pri_enabled))return -EBUSY ;
@@ -246,6 +269,14 @@ void pci_disable_pri(struct pci_dev *pdev) {u16 control; +/* +* As per PCIe r4.0, sec 9.3.7.11, only PF is permitted to +* implement PRI and all associated VFs can only use it. +* So don't do anything for VF and just return. +*/ +if (pdev->is_virtfn) +return ; +if (WARN_ON (!pdev->pri_enabled))return ;
@@ -269,6 +300,9 @@ void pci_restore_pri_state(struct pci_dev *pdev)u16 control =PCI_PRI_CTRL_ENABLE ;u32 reqs = pdev->pri_reqs_alloc; +if (pdev->is_virtfn) +return ; +if (!pdev->pri_enabled)return ;
@@ -291,6 +325,9 @@ int pci_reset_pri(struct pci_dev *pdev) {u16 control; +if (pdev->is_virtfn) +return 0 ; +if (WARN_ON (pdev->pri_enabled))return -EBUSY ;
@@ -427,11 +464,12 @@ EXPORT_SYMBOL_GPL(pci_pasid_features);int pci_prg_resp_pasid_required (struct pci_dev *pdev) {u16 status; +struct pci_dev *pf =pci_physfn (pdev); -if (!pdev->pri_cap) +if (!pf->pri_cap)return 0 ; -pci_read_config_word (pdev, pdev->pri_cap +PCI_PRI_STATUS , &status); +pci_read_config_word (pf, pf->pri_cap +PCI_PRI_STATUS , &status);if (status &PCI_PRI_STATUS_PASID )return 1 ;
Signature · 1 line
2.21.0
PATCH v7 5/7 PCI/ATS: Add PASID support for PCIe VF devices
In reply to [email protected]From: Kuppuswamy Sathyanarayanan <[email protected]>
When IOMMU tries to enable PASID for VF device in iommu_enable_dev_iotlb(), it always fails because PASID support for PCIe VF device is currently broken in PCIE driver. Current implementation expects the given PCIe device (PF & VF) to implement PASID capability before enabling the PASID support. But this assumption is incorrect. As per PCIe spec r4.0, sec 9.3.7.14, all VFs associated with PF can only use the PASID of the PF and not implement it. Hence we need to create exception for handling the PASID support for PCIe VF device.
--- drivers/pci/ats.c | 49 +++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 43 insertions(+), 6 deletions(-)
drivers/pci/ats.c +43 -6
@@ -43,6 +43,15 @@ static void pci_pasid_init(struct pci_dev *pdev)#ifdef CONFIG_PCI_PASID int pos; +/* +* As per PCIe r4.0, sec 9.3.7.14, only PF is permitted to +* implement PASID Capability and all associated VFs can +* only use it. Since PF already initialized the PASID +* parameters there is no need to proceed further. +*/ +if (pdev->is_virtfn) +return ; + pos =pci_find_ext_capability (pdev,PCI_EXT_CAP_ID_PASID );if (!pos)return ;
@@ -355,7 +364,20 @@ EXPORT_SYMBOL_GPL(pci_reset_pri);int pci_enable_pasid (struct pci_dev *pdev,int features) {u16 control, supported; +struct pci_dev *pf =pci_physfn (pdev); +/* +* IOMMU is the only user of this function and as per +* current usage, PF PASID enable always happens before +* VF and hence we don't need to do anything special +* for VF. So just return success if PASID is enabled in PF. +*/ +if (pdev->is_virtfn) { +if (pf->pasid_enabled) +return 0 ; +else +return -EINVAL ; + }if (WARN_ON (pdev->pasid_enabled))return -EBUSY ;
@@ -392,6 +414,14 @@ void pci_disable_pasid(struct pci_dev *pdev) {u16 control =0 ; +/* +* As per PCIe r4.0, sec 9.3.7.14, only PF is permitted to +* implement PASID Capability and all associated VFs can +* only use it. So don't do anything for VF and just return. +*/ +if (pdev->is_virtfn) +return ; +if (WARN_ON (!pdev->pasid_enabled))return ;
@@ -412,6 +442,13 @@ void pci_restore_pasid_state(struct pci_dev *pdev) {u16 control; +/* +* PF should have already restored the PASID state. So for +* VF, just return. +*/ +if (pdev->is_virtfn) +return ; +if (!pdev->pasid_enabled)return ;
@@ -436,12 +473,12 @@ EXPORT_SYMBOL_GPL(pci_restore_pasid_state);int pci_pasid_features (struct pci_dev *pdev) {u16 supported; +struct pci_dev *pf =pci_physfn (pdev); -if (!pdev->pasid_cap) +if (!pf->pasid_cap)return -EINVAL ; -pci_read_config_word (pdev, pdev->pasid_cap +PCI_PASID_CAP , - &supported); +pci_read_config_word (pf, pf->pasid_cap +PCI_PASID_CAP , &supported); supported &=PCI_PASID_CAP_EXEC |PCI_PASID_CAP_PRIV ;
@@ -492,12 +529,12 @@ EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);int pci_max_pasids (struct pci_dev *pdev) {u16 supported; +struct pci_dev *pf =pci_physfn (pdev); -if (!pdev->pasid_cap) +if (!pf->pasid_cap)return -EINVAL ; -pci_read_config_word (pdev, pdev->pasid_cap +PCI_PASID_CAP , - &supported); +pci_read_config_word (pf, pf->pasid_cap +PCI_PASID_CAP , &supported); supported = (supported &PASID_NUMBER_MASK ) >>PASID_NUMBER_SHIFT ;
Signature · 1 line
2.21.0
PATCH v7 6/7 PCI/ATS: Disable PF/VF ATS service independently
In reply to [email protected]From: Kuppuswamy Sathyanarayanan <[email protected]>
Currently all VF's needs to be disable their ATS service before disabling the ATS service in corresponding PF device. But this logic is incorrect and does not align with the spec. Also it might lead to some power and performance impact in the system. As per PCIe spec r4.0, sec 9.3.7.8, ATS Capabilities in VFs and their associated PFs may be enabled/disabled independently. So remove this dependency logic in enable/disable code.
--- drivers/pci/ats.c | 11 ----------- include/linux/pci.h | 1 - 2 files changed, 12 deletions(-)
drivers/pci/ats.c +0 -11
@@ -108,8 +108,6 @@ int pci_enable_ats(struct pci_dev *dev, int ps) pdev =pci_physfn (dev);if (pdev->ats_stu != ps)return -EINVAL ; - -atomic_inc (&pdev->ats_ref_cnt);/* count enabled VFs */ }else { dev->ats_stu = ps; ctrl |=PCI_ATS_CTRL_STU (dev->ats_stu -PCI_ATS_MIN_STU );
@@ -127,20 +125,11 @@ EXPORT_SYMBOL_GPL(pci_enable_ats);*/ void pci_disable_ats (struct pci_dev *dev) { -struct pci_dev *pdev;u16 ctrl;if (WARN_ON (!dev->ats_enabled))return ; -if (atomic_read (&dev->ats_ref_cnt)) -return ;/* VFs still enabled */ - -if (dev->is_virtfn) { - pdev =pci_physfn (dev); -atomic_dec (&pdev->ats_ref_cnt); - } -pci_read_config_word (dev, dev->ats_cap +PCI_ATS_CTRL , &ctrl); ctrl &= ~PCI_ATS_CTRL_ENABLE ;pci_write_config_word (dev, dev->ats_cap +PCI_ATS_CTRL , ctrl);
include/linux/pci.h +0 -1
@@ -452,7 +452,6 @@ struct pci_dev { };u16 ats_cap;/* ATS Capability offset */ u8 ats_stu;/* ATS Smallest Translation Unit */ -atomic_t ats_ref_cnt;/* Number of VFs with ATS enabled */ #endif #ifdef CONFIG_PCI_PRI u16 pri_cap;/* PRI Capability offset */
Signature · 1 line
2.21.0
PATCH v7 7/7 PCI: Skip Enhanced Allocation (EA) initialization for VF device
In reply to [email protected]From: Kuppuswamy Sathyanarayanan <[email protected]>
As per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced Allocation Capability. So skip pci_ea_init() for virtual devices.
--- drivers/pci/pci.c | 7 +++++++ 1 file changed, 7 insertions(+)
drivers/pci/pci.c +7 -0
@@ -3025,6 +3025,13 @@ void pci_ea_init(struct pci_dev *dev)int offset;int i; +/* +* Per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced +* Allocation Capability. +*/ +if (dev->is_virtfn) +return ; +/* find PCI EA capability in list */ ea =pci_find_capability (dev,PCI_CAP_ID_EA );if (!ea)
Signature · 1 line
2.21.0
Re: [PATCH v7 7/7] PCI: Skip Enhanced Allocation (EA) initialization for VF device
Hi Bjorn,
Quoted 9 lines · On 8/28/19 3:14 PM, [email protected] wrote:
From: Kuppuswamy Sathyanarayanan <[email protected]>
As per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced Allocation Capability. So skip
pci_ea_init()for virtual devices.Cc Ashok Raj <[email protected]>Cc Keith Busch <[email protected]>Suggested-by Ashok Raj <[email protected]>Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>
This patch was also dropped in your v8. Is this also intentional?
Quoted 22 lines
--- drivers/pci/pci.c | 7 +++++++ 1 file changed, 7 insertions(+)drivers/pci/pci.c +7
-0@@ -3025,6 +3025,13 @@ void pci_ea_init(struct pci_dev *dev)int offset;int i; +/* +* Per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced +* Allocation Capability. +*/ +if (dev->is_virtfn) +return ; +/* find PCI EA capability in list */ ea =pci_find_capability (dev,PCI_CAP_ID_EA );if (!ea)
Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 7/7] PCI: Skip Enhanced Allocation (EA) initialization for VF device
Quoted 5 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Quoted 2 lines · On 8/28/19 3:14 PM, [email protected] wrote:
From: Kuppuswamy Sathyanarayanan <[email protected]>
Quoted 8 lines
Quoted 7 lines
As per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced Allocation Capability. So skip
pci_ea_init()for virtual devices.Cc Ashok Raj <[email protected]>Cc Keith Busch <[email protected]>Suggested-by Ashok Raj <[email protected]>Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>This patch was also dropped in your v8. Is this also intentional?
Yes, I dropped it because I didn't think there was much motivation for it.
If a device is broken, i.e., a VF has an EA capability, this patch silently returns. The existing code would try to use the EA capability and something would probably blow up, so in that sense, this patch makes the hardware issue less visible.
If a device is correct, i.e., a VF does *not* have an EA capability, pci_find_capability() will fail anyway, so this patch doesn't change the functional behavior.
This patch *does* avoid the pci_find_capability() in that case, which is a performance optimization. We could merge it on that basis, but we should try to quantify the benefit to see if it's really worthwhile and the commit log should use that as the explicit motivation.
Quoted 26 lines
Quoted 21 lines
--- drivers/pci/pci.c | 7 +++++++ 1 file changed, 7 insertions(+)drivers/pci/pci.c +7
-0@@ -3025,6 +3025,13 @@ void pci_ea_init(struct pci_dev *dev)int offset;int i; +/* +* Per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced +* Allocation Capability. +*/ +if (dev->is_virtfn) +return ; +/* find PCI EA capability in list */ ea =pci_find_capability (dev,PCI_CAP_ID_EA );if (!ea)Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Re: [PATCH v7 7/7] PCI: Skip Enhanced Allocation (EA) initialization for VF device
Quoted 9 lines · Bjorn Helgaas
Quoted 8 lines · Kuppuswamy Sathyanarayanan
Hi Bjorn,
Quoted 5 lines · On 8/28/19 3:14 PM, [email protected] wrote:
From: Kuppuswamy Sathyanarayanan <[email protected]>
As per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced Allocation Capability. So skip
pci_ea_init()for virtual devices.
Quoted 8 lines
Quoted 5 lines
Quoted 4 lines
Cc Ashok Raj <[email protected]>Cc Keith Busch <[email protected]>Suggested-by Ashok Raj <[email protected]>Signed-off-by Kuppuswamy Sathyanarayanan <[email protected]>This patch was also dropped in your v8. Is this also intentional?
Yes, I dropped it because I didn't think there was much motivation for it.
Agreed!
Quoted 9 lines
If a device is broken, i.e., a VF has an EA capability, this patch silently returns. The existing code would try to use the EA capability and something would probably blow up, so in that sense, this patch makes the hardware issue less visible.
If a device is correct, i.e., a VF does *not* have an EA capability,
pci_find_capability()will fail anyway, so this patch doesn't change the functional behavior.
But do you think while at this can we atleast do a warning to make sure HW probably messed up just after the EA capability is read? Atleast it would be an early warning vs. it trying to break later. Like the other issues we ran into with the PIN interrupt accidently set in some hardware for VF's for instance.
Quoted 32 lines
This patch *does* avoid the
pci_find_capability()in that case, which is a performance optimization. We could merge it on that basis, but we should try to quantify the benefit to see if it's really worthwhile and the commit log should use that as the explicit motivation.Quoted 26 lines
Quoted 21 lines
--- drivers/pci/pci.c | 7 +++++++ 1 file changed, 7 insertions(+)drivers/pci/pci.c +7
-0@@ -3025,6 +3025,13 @@ void pci_ea_init(struct pci_dev *dev)int offset;int i; +/* +* Per PCIe r4.0, sec 9.3.6, VF must not implement Enhanced +* Allocation Capability. +*/ +if (dev->is_virtfn) +return ; +/* find PCI EA capability in list */ ea =pci_find_capability (dev,PCI_CAP_ID_EA );if (!ea)Signature · 2 lines
Sathyanarayanan Kuppuswamy Linux kernel developer
Quoted 33 lines · On Wed, Aug 28, 2019 at 03:14:00PM -0700…
From: Kuppuswamy Sathyanarayanan <[email protected]>
Current implementation of ATS, PASID, PRI does not handle VF dependencies correctly. Following patches addresses this issue.
Changes since v6:
* Removed locking interfaces used in v6.
* Made necessary changes to PRI/PASID enable/disable calls to allow register changes only for PF.Changes since v5:
* Created new patches for PRI/PASID capability caching.
* Removed individual locks (pri_lock,pasid_lock) and added common resource lock to synchronize PRI/PASID updates between PF/VF.
* Addressed comments from Bjorn Helgaas.Changes since v4:
* Defined empty functions forpci_pri_init()andpci_pasid_init()for cases whereCONFIG_PCI_PRIandCONFIG_PCI_PASIDare not enabled.Changes since v3:
* Fixed critical path (lock context) in pci_restore_*_state functions.Changes since v2:
* Added locking mechanism to synchronize accessing PF registers in VF.
* Removed spec compliance checks in patches.
* Addressed comments from Bjorn Helgaas.Changes since v1:
* Added more details about the patches in commit log.
* Removed bulk spec check patch.
* Addressed comments from Bjorn Helgaas.
Quoted 8 lines
Kuppuswamy Sathyanarayanan (7):
PCI/ATS: Fixpci_prg_resp_pasid_required()dependency issues PCI/ATS: Cache PRI capability check result
PCI/ATS: Cache PASID capability check result
PCI/ATS: Add PRI support for PCIe VF devices
PCI/ATS: Add PASID support for PCIe VF devices PCI/ATS: Disable PF/VF ATS service independently
PCI: Skip Enhanced Allocation (EA) initialization for VF device
To make it easier to backport things, I think these should be reordered so the important fixes are first, e.g., like this:
PCI/ATS: Add PRI support for PCIe VF devices PCI/ATS: Add PASID support for PCIe VF devices PCI/ATS: Disable PF/VF ATS service independently PCI/ATS: Cache PRI capability check result PCI/ATS: Cache PASID capability check result
I don't think the following ones are actually needed:
PCI: Skip Enhanced Allocation (EA) initialization for VF device PCI/ATS: Fix pci_prg_resp_pasid_required() dependency issues
I'll post a v8 with my proposal.
Quoted 9 lines
drivers/pci/ats.c | 202 ++++++++++++++++++++++++++++------------ drivers/pci/pci.c | 7 ++ include/linux/pci-ats.h | 12 ++- include/linux/pci.h | 3 +- 4 files changed, 159 insertions(+), 65 deletions(-)Signature · 1 line
2.21.0