PCI/ASPM: Mask ASPM states based on Devicetree properties
v6 cover PCI/ASPM: Mask ASPM states based on Devicetree properties
---
Changes in v6:
- Disable/enable each function L0s/L1. (sashiko)
- Move from pcie_capability_write_word to pcie_capability_clear_and_set_word (Sashiko)
- Link to v5: https://patch.msgid.link/[email protected]
Changes in v5:
- update all the child devices when enabling the ASPM back (sashiko).
- Link to v4: https://patch.msgid.link/[email protected]
Changes in v4:
- Clear link->aspm_enabled alongside aspm_support/aspm_default when an ASPM state is masked, so it no longer goes stale relative to hardware (sashiko).
- Mask the LNKCTL restore snapshot against the post-override
aspm_support so the restore step can no longer re-enable a state the Devicetree override just disabled. (sashiko)
- Link to v3: https://patch.msgid.link/[email protected]
Changes in v3:
- Move pcie_aspm_override_default_link_state() call in pcie_aspm_cap_init() to before the "Restore L0s/L1" block. pcie_aspm_cap_init() disables L1 in hardware prior to aspm_l1ss_init() and re-enables it only in the restore block. Calling pcie_config_aspm_l1ss() while L1 is already disabled satisfies its precondition ("Caller must disable L1 first"), whereas the previous placement after the restore violated it (sashiko).
- Link to v2: https://patch.msgid.link/[email protected]
Changes in v2:
- Disable L1ss when L1 is disabled as pointed by sashiko.
- Disable L1ss if bootloader enables them but we are disabling via devicetree pointed by sashiko.
- Link to v1: https://patch.msgid.link/[email protected]
---
Krishna Chaitanya Chundru (3):
PCI/ASPM: Use pcie_capability_clear_and_set_word() for ASPM disable/restore PCI/ASPM: Disable/restore ASPM on every function for multi-function devices PCI/ASPM: Mask ASPM states based on Devicetree properties
drivers/pci/pcie/aspm.c | 149 +++++++++++++++++++++++++++++++++--------------- 1 file changed, 102 insertions(+), 47 deletions(-)
---
base-commit: f5098b6bae761e346ebcd9da7f95622c04733cff
change-id: 20260429-aspm-da55a54279e4
Best regards,
Signature · 1 line
Krishna Chaitanya Chundru <[email protected]>
PATCH v6 1/3 PCI/ASPM: Use pcie_capability_clear_and_set_word() for ASPM disable/restore
pcie_aspm_cap_init() disables ASPM L0s/L1 on both ends of the Link before touching L1SS config, then later restores the LNKCTL state that was in effect beforehand. Both steps use raw pcie_capability_write_word() calls: the disable step computes the new value by hand from a snapshot taken earlier in the function, and the restore step writes that same snapshot straight back.
Switch both steps to pcie_capability_clear_and_set_word(), masked to PCI_EXP_LNKCTL_ASPMC, matching the accessor pcie_config_aspm_dev() already uses elsewhere in this file for the exact same register. This does a live read-modify-write of just the ASPM Control bits instead of relying on a stale snapshot for the rest of the word, and is consistent with how the rest of the file already touches this register. No functional change.
--- drivers/pci/pcie/aspm.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-)
drivers/pci/pcie/aspm.c +10 -6
@@ -894,10 +894,10 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)/* Disable L0s/L1 before updating L1SS config */ if (FIELD_GET (PCI_EXP_LNKCTL_ASPMC , child_lnkctl) ||FIELD_GET (PCI_EXP_LNKCTL_ASPMC , parent_lnkctl)) { -pcie_capability_write_word (child,PCI_EXP_LNKCTL , - child_lnkctl & ~PCI_EXP_LNKCTL_ASPMC ); -pcie_capability_write_word (parent,PCI_EXP_LNKCTL , - parent_lnkctl & ~PCI_EXP_LNKCTL_ASPMC ); +pcie_capability_clear_and_set_word (child,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC ,0 ); +pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC ,0 ); }/*
@@ -927,8 +927,12 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)/* Restore L0s/L1 if they were enabled */ if (FIELD_GET (PCI_EXP_LNKCTL_ASPMC , child_lnkctl) ||FIELD_GET (PCI_EXP_LNKCTL_ASPMC , parent_lnkctl)) { -pcie_capability_write_word (parent,PCI_EXP_LNKCTL , parent_lnkctl); -pcie_capability_write_word (child,PCI_EXP_LNKCTL , child_lnkctl); +pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC , + parent_lnkctl &PCI_EXP_LNKCTL_ASPMC ); +pcie_capability_clear_and_set_word (child,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC , + child_lnkctl &PCI_EXP_LNKCTL_ASPMC ); }/* Save default state */
Signature · 1 line
2.34.1
[email protected]
AI reviewRe: [PATCH v6 1/3] PCI/ASPM: Use pcie_capability_clear_and_set_word() for ASPM disable/restore
Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 15 lines
pcie_aspm_cap_init()disables ASPM L0s/L1 on both ends of the Link before touching L1SS config, then later restores the LNKCTL state that was in effect beforehand. Both steps use rawpcie_capability_write_word()calls: the disable step computes the new value by hand from a snapshot taken earlier in the function, and the restore step writes that same snapshot straight back.Switch both steps to
pcie_capability_clear_and_set_word(), masked toPCI_EXP_LNKCTL_ASPMC, matching the accessorpcie_config_aspm_dev()already uses elsewhere in this file for the exact same register. This does a live read-modify-write of just the ASPM Control bits instead of relying on a stale snapshot for the rest of the word, and is consistent with how the rest of the file already touches this register. No functional change.
Quoted 2 lines
Fixes 7447990137bf ("PCI/ASPM: Disable L1 before disabling L1 PM Substates")Signed-off-by Krishna Chaitanya Chundru <[email protected]>
Sashiko has reviewed this patch and found no issues. It looks great!
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Re: [PATCH v6 1/3] PCI/ASPM: Use pcie_capability_clear_and_set_word() for ASPM disable/restore
In reply to Krishna Chaitanya ChundruQuoted 15 lines · Krishna Chaitanya Chundru
pcie_aspm_cap_init()disables ASPM L0s/L1 on both ends of the Link before touching L1SS config, then later restores the LNKCTL state that was in effect beforehand. Both steps use rawpcie_capability_write_word()calls: the disable step computes the new value by hand from a snapshot taken earlier in the function, and the restore step writes that same snapshot straight back.Switch both steps to
pcie_capability_clear_and_set_word(), masked toPCI_EXP_LNKCTL_ASPMC, matching the accessorpcie_config_aspm_dev()already uses elsewhere in this file for the exact same register. This does a live read-modify-write of just the ASPM Control bits instead of relying on a stale snapshot for the rest of the word, and is consistent with how the rest of the file already touches this register. No functional change.
Quoted 1 line
Fixes 7447990137bf ("PCI/ASPM: Disable L1 before disabling L1 PM Substates")
Quoted 1 line
Signed-off-by Krishna Chaitanya Chundru <[email protected]>
- Mani
Signature · 1 line
மணிவண்ணன் சதாசிவம்
PATCH v6 2/3 PCI/ASPM: Disable/restore ASPM on every function for multi-function devices
In reply to Krishna Chaitanya Chundrupcie_aspm_cap_init() disables ASPM L0s/L1 before touching L1SS config, then restores the pre-existing state afterward. Both steps only ever touched link->downstream, i.e. function 0 of the downstream component, leaving sibling functions (>0) on a multi-function device untouched.
This means the "disable" step does not actually disable ASPM link-wide on a multi-function device: a sibling function can still have L1 enabled even after this step runs. PCIe r6.2 sec 7.5.3.7 recommends programming the same ASPM Control value for all functions of a multi-function device, and pcie_config_aspm_link() already loops over every function on the bus for exactly this reason.
Loop over every function on linkbus->devices for both the disable and restore steps, keeping the existing PCIe r6.2 sec 7.5.3.7 ordering (disable downstream functions before upstream, restore upstream before downstream functions). The masked pcie_capability_clear_and_set_word() accessor from the previous commit makes this safe: it only ever touches the ASPM Control bits, so function-specific bits elsewhere in LNKCTL (e.g. Read Completion Boundary, CLKREQ Enable) on sibling functions are left untouched.
--- drivers/pci/pcie/aspm.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
drivers/pci/pcie/aspm.c +8 -5
@@ -861,6 +861,7 @@ static void pcie_aspm_override_default_link_state(struct pcie_link_state *link)static void pcie_aspm_cap_init (struct pcie_link_state *link,int blacklist) {struct pci_dev *child = link->downstream, *parent = link->pdev; +struct pci_dev *fn;u16 parent_lnkctl, child_lnkctl;struct pci_bus *linkbus = parent->subordinate;
@@ -894,8 +895,9 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)/* Disable L0s/L1 before updating L1SS config */ if (FIELD_GET (PCI_EXP_LNKCTL_ASPMC , child_lnkctl) ||FIELD_GET (PCI_EXP_LNKCTL_ASPMC , parent_lnkctl)) { -pcie_capability_clear_and_set_word (child,PCI_EXP_LNKCTL , -PCI_EXP_LNKCTL_ASPMC ,0 ); +list_for_each_entry (fn, &linkbus->devices, bus_list) +pcie_capability_clear_and_set_word (fn,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC ,0 );pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL ,PCI_EXP_LNKCTL_ASPMC ,0 ); }
@@ -930,9 +932,10 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL ,PCI_EXP_LNKCTL_ASPMC , parent_lnkctl &PCI_EXP_LNKCTL_ASPMC ); -pcie_capability_clear_and_set_word (child,PCI_EXP_LNKCTL , -PCI_EXP_LNKCTL_ASPMC , - child_lnkctl &PCI_EXP_LNKCTL_ASPMC ); +list_for_each_entry (fn, &linkbus->devices, bus_list) +pcie_capability_clear_and_set_word (fn,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC , + child_lnkctl &PCI_EXP_LNKCTL_ASPMC ); }/* Save default state */
Signature · 1 line
2.34.1
[email protected]
AI reviewRe: [PATCH v6 2/3] PCI/ASPM: Disable/restore ASPM on every function for multi-function devices
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:
Pre-existing issues:
- [High] The conditional check guarding the ASPM disable/restore sequence evaluates only function 0, bypassing the new multi-function loop if only a sibling function has ASPM enabled.
- [High] L1 PM Substates (L1SS) configuration is only applied to function 0, violating PCIe specifications for multi-function devices.
- [High] NULL pointer dereference in pcie_aspm_cap_init if a PCIe link has downstream devices but no function 0.
- [High] Spec violation in pci_restore_aspm_l1ss_state() where ASPMC is temporarily disabled only on the parent and the active function, ignoring sibling functions. --
--- Patch [2]: [PATCH v6 2/3] PCI/ASPM: Disable/restore ASPM on every function for multi-function devices --- commit b996c9e712ad7161b9278ed4aa901d8d723bf24a
Author: Krishna Chaitanya Chundru <[email protected]>
PCI/ASPM: Disable/restore ASPM on every function for multi-function devices
This patch modifies pcie_aspm_cap_init() to disable and restore ASPM L0s/L1 states across all functions of a multi-function device. It replaces the direct configuration of only the downstream function 0 with a loop over linkbus->devices.
drivers/pci/pcie/aspm.c static void pcie_aspm_override_default_link_state(struct pcie_link_state *link) · line 864
Quoted 9 lines
drivers/pci/pcie/aspm.c +1
-0@@ -861,6 +861,7 @@ static void pcie_aspm_override_default_link_state(struct pcie_link_state *link)static void pcie_aspm_cap_init (struct pcie_link_state *link,int blacklist) {struct pci_dev *child = link->downstream, *parent = link->pdev; +struct pci_dev *fn;
[Severity: High]
This is a pre-existing issue, but is it possible for child to be NULL here and dereferenced later in this function?
If a PCIe bus has downstream devices but lacks function 0 (for example, due to dynamic sysfs removal, SR-IOV configurations, or device passthrough), pci_function_0() returns NULL. This causes alloc_pcie_link_state() to set link->downstream to NULL.
Later in pcie_aspm_cap_init(), child is unconditionally dereferenced:
pcie_aspm_cap_init() {
...
if (!(parent->aspm_l0s_support && child->aspm_l0s_support) &&
!(parent->aspm_l1_support && child->aspm_l1_support))
return;
...
}Can this lead to a NULL pointer dereference panic?
[ ... ]
drivers/pci/pcie/aspm.c static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist) · line 900
Quoted 9 lines
drivers/pci/pcie/aspm.c +3
-2@@ -894,8 +895,9 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)/* Disable L0s/L1 before updating L1SS config */ if (FIELD_GET (PCI_EXP_LNKCTL_ASPMC , child_lnkctl) ||FIELD_GET (PCI_EXP_LNKCTL_ASPMC , parent_lnkctl)) { -pcie_capability_clear_and_set_word (child,PCI_EXP_LNKCTL , -PCI_EXP_LNKCTL_ASPMC ,0 ); +list_for_each_entry (fn, &linkbus->devices, bus_list) +pcie_capability_clear_and_set_word (fn,PCI_EXP_LNKCTL , +PCI_EXP_LNKCTL_ASPMC ,0 );
[Severity: High]
This is a pre-existing issue, but does this conditional check bypass the new multi-function loop if only a sibling function has ASPM enabled?
The if condition evaluates only child_lnkctl (function 0) and parent_lnkctl. If a multi-function device is initialized where function 0 and the parent port have ASPM L0s/L1 disabled, but a sibling function has it enabled, it appears this check evaluates to false.
This would completely bypass the newly added loop, allowing L1SS configuration to proceed while a sibling function still has ASPM L1 enabled, violating the PCIe spec.
drivers/pci/pcie/aspm.c static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist) · line 903
Quoted 3 lines
drivers/pci/pcie/aspm.c +0
-0pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL ,PCI_EXP_LNKCTL_ASPMC ,0 ); }
[ ... ]
[Severity: High]
This is a pre-existing issue, but does pcie_config_aspm_l1ss() need a similar multi-function loop?
Looking at pcie_config_aspm_l1ss() in drivers/pci/pcie/aspm.c:
pcie_config_aspm_l1ss() {
...
/* Disable all L1 substates */
pci_clear_and_set_config_dword(child, child->l1ss + PCI_L1SS_CTL1,
PCI_L1SS_CTL1_L1SS_MASK, 0);
pci_clear_and_set_config_dword(parent, parent->l1ss + PCI_L1SS_CTL1,
PCI_L1SS_CTL1_L1SS_MASK, 0);
...
}It appears L1SS configuration is only applied to child (function 0) and the parent. Does this leave L1 PM Substates out of sync across sibling functions, violating the PCIe r6.2 sec 7.8.3.3 mandate for identical L1SS programming across all functions of a multi-function device?
[Severity: High]
This is a pre-existing issue, but is there a similar multi-function ordering violation in pci_restore_aspm_l1ss_state()?
Looking at pci_restore_aspm_l1ss_state() in drivers/pci/pcie/aspm.c:
pci_restore_aspm_l1ss_state() {
...
if (FIELD_GET(PCI_EXP_LNKCTL_ASPMC, clnkctl) ||
FIELD_GET(PCI_EXP_LNKCTL_ASPMC, plnkctl)) {
pcie_capability_write_word(pdev, PCI_EXP_LNKCTL,
clnkctl & ~PCI_EXP_LNKCTL_ASPMC);
pcie_capability_write_word(parent, PCI_EXP_LNKCTL,
plnkctl & ~PCI_EXP_LNKCTL_ASPMC);
}
...
}When called on a multi-function device for function 1, it temporarily disables ASPMC on function 1 and immediately on the parent. If function 0 still has ASPMC enabled, the upstream parent is disabled before the downstream function 0.
Does this violate the PCIe spec requirement to disable downstream functions before upstream functions, potentially causing link drops during device resume?
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
Re: [PATCH v6 2/3] PCI/ASPM: Disable/restore ASPM on every function for multi-function devices
In reply to Krishna Chaitanya ChundruQuoted 21 lines · Krishna Chaitanya Chundru
pcie_aspm_cap_init()disables ASPM L0s/L1 before touching L1SS config, then restores the pre-existing state afterward. Both steps only ever touchedlink->downstream, i.e. function 0 of the downstream component, leaving sibling functions (>0) on a multi-function device untouched.This means the "disable" step does not actually disable ASPM link-wide on a multi-function device: a sibling function can still have L1 enabled even after this step runs. PCIe r6.2 sec 7.5.3.7 recommends programming the same ASPM Control value for all functions of a multi-function device, and
pcie_config_aspm_link()already loops over every function on the bus for exactly this reason.Loop over every function on
linkbus->devicesfor both the disable and restore steps, keeping the existing PCIe r6.2 sec 7.5.3.7 ordering (disable downstream functions before upstream, restore upstream before downstream functions). The maskedpcie_capability_clear_and_set_word()accessor from the previous commit makes this safe: it only ever touches the ASPM Control bits, so function-specific bits elsewhere in LNKCTL (e.g. Read Completion Boundary, CLKREQ Enable) on sibling functions are left untouched.
Quoted 1 line
Fixes 7447990137bf ("PCI/ASPM: Disable L1 before disabling L1 PM Substates")
Quoted 1 line
Signed-off-by Krishna Chaitanya Chundru <[email protected]>
- Mani
Signature · 1 line
மணிவண்ணன் சதாசிவம்
PATCH v6 3/3 PCI/ASPM: Mask ASPM states based on Devicetree properties
In reply to Krishna Chaitanya ChundruSome platforms require selectively disabling specific ASPM states on a given PCIe link to avoid link instability or functional failures caused by board-level connectivity constraints such as PCB routing, connectors, slots, or external cabling.
Devicetree supports disabling ASPM L0s, L1, and L1 PM Substates via the 'aspm-no-l0s', 'aspm-no-l1' [1], and 'aspm-no-l1ss' [2] properties. However, the ASPM driver does not currently honor these properties when initializing the default link state.
When firmware enables L1 PM Substates before the kernel takes over, masking aspm_support alone is insufficient to disable them in hardware. pcie_config_aspm_link() guards L1SS configuration behind a check on aspm_capable, which is derived from aspm_support. Once aspm_support is masked, pcie_config_aspm_l1ss() is never called, leaving firmware-enabled L1SS substates active in hardware.
Fix this by introducing pcie_link_has_aspm_override() to check for DT override properties on either endpoint of the link. In pcie_aspm_override_default_link_state(), use it to:
- Mask aspm_support, aspm_default, and aspm_enabled for any disabled state, so software's view of the link stays in sync with what is actually programmed in hardware. Leaving aspm_enabled stale would make pcie_aspm_enabled() and the aspm sysfs attributes report a state as active even after it has been masked, and could cause pcie_config_aspm_link()'s "already in requested state" check to skip reprogramming hardware to match.
- Explicitly call pcie_config_aspm_l1ss(link, 0) before masking aspm_support when firmware has L1SS active and DT requests disabling L1 or L1SS, since pcie_config_aspm_link() will no longer do so once aspm_capable is derived from the masked aspm_support.
Move the aspm_default initialization and pcie_aspm_override_default_link_state() call in pcie_aspm_cap_init() to before the "Restore L0s/L1" block. pcie_aspm_cap_init() disables L1 in hardware prior to aspm_l1ss_init() and re-enables it only in the restore block. Calling pcie_config_aspm_l1ss() while L1 is already disabled satisfies its precondition ("Caller must disable L1 first"), whereas the previous placement after the restore violated it.
Since the restore block writes back the parent_lnkctl/child_lnkctl snapshot taken from hardware before the DT override ran, mask the L0s and L1 enable bits out of that snapshot for any state the override has just disabled in aspm_support. Otherwise the restore step would unconditionally reprogram the link back to firmware's original L0s/L1 configuration, defeating the Devicetree override it is meant to enforce.
Move pcie_config_aspm_l1ss() earlier in the file so it can be called from pcie_aspm_override_default_link_state().
Link [1]: https://github.com/devicetree-org/dt-schema/pull/188
Link [2]: https://github.com/devicetree-org/dt-schema/pull/190
--- drivers/pci/pcie/aspm.c | 132 +++++++++++++++++++++++++++++++++--------------- 1 file changed, 90 insertions(+), 42 deletions(-)
drivers/pci/pcie/aspm.c +90 -42
@@ -839,6 +839,49 @@ static void aspm_l1ss_init(struct pcie_link_state *link)#define FLAG (x, y, d) (((x) & (PCIE_LINK_STATE_##y)) ? d : "") +/* Configure the ASPM L1 substates. Caller must disable L1 first. */ +static void pcie_config_aspm_l1ss (struct pcie_link_state *link,u32 state) +{ +u32 val =0 ; +struct pci_dev *child = link->downstream, *parent = link->pdev; + +if (state &PCIE_LINK_STATE_L1_1 ) + val |=PCI_L1SS_CTL1_ASPM_L1_1 ; +if (state &PCIE_LINK_STATE_L1_2 ) + val |=PCI_L1SS_CTL1_ASPM_L1_2 ; +if (state &PCIE_LINK_STATE_L1_1_PCIPM ) + val |=PCI_L1SS_CTL1_PCIPM_L1_1 ; +if (state &PCIE_LINK_STATE_L1_2_PCIPM ) + val |=PCI_L1SS_CTL1_PCIPM_L1_2 ; + +/* +* PCIe r6.2, sec 5.5.4, rules for enabling L1 PM Substates: +* - Clear L1.x enable bits at child first, then at parent +* - Set L1.x enable bits at parent first, then at child +* - ASPM/PCIPM L1.2 must be disabled while programming timing +* parameters +*/ + +/* Disable all L1 substates */ +pci_clear_and_set_config_dword (child, child->l1ss +PCI_L1SS_CTL1 , +PCI_L1SS_CTL1_L1SS_MASK ,0 ); +pci_clear_and_set_config_dword (parent, parent->l1ss +PCI_L1SS_CTL1 , +PCI_L1SS_CTL1_L1SS_MASK ,0 ); + +/* Enable what we need to enable */ +pci_clear_and_set_config_dword (parent, parent->l1ss +PCI_L1SS_CTL1 , +PCI_L1SS_CTL1_L1SS_MASK , val); +pci_clear_and_set_config_dword (child, child->l1ss +PCI_L1SS_CTL1 , +PCI_L1SS_CTL1_L1SS_MASK , val); +} + +static bool pcie_link_has_aspm_override (const struct pcie_link_state *link, +const char *aspm) +{ +return (device_property_present (&link->pdev->dev, aspm) || +device_property_present (&link->downstream->dev, aspm)); +} +static void pcie_aspm_override_default_link_state (struct pcie_link_state *link) {struct pci_dev *pdev = link->downstream;
@@ -846,6 +889,36 @@ static void pcie_aspm_override_default_link_state(struct pcie_link_state *link)/* For devicetree platforms, enable L0s and L1 by default */ if (of_have_populated_dt ()) { +bool no_l0s =pcie_link_has_aspm_override (link,"aspm-no-l0s" ); +bool no_l1 =pcie_link_has_aspm_override (link,"aspm-no-l1" ); +bool no_l1ss =pcie_link_has_aspm_override (link,"aspm-no-l1ss" ); + +if (no_l0s) { + link->aspm_support &= ~PCIE_LINK_STATE_L0S ; + link->aspm_default &= ~PCIE_LINK_STATE_L0S ; + link->aspm_enabled &= ~PCIE_LINK_STATE_L0S ; + } + +/* +* Clear L1SS in hardware before updating aspm_support. Once +* aspm_capable is derived from aspm_support, pcie_config_aspm_link() +* skips pcie_config_aspm_l1ss() entirely via the aspm_capable guard, +* leaving firmware-enabled L1SS substates active in hardware. +* This applies equally when disabling L1 (which implies L1SS). +*/ +if ((no_l1 || no_l1ss) && (link->aspm_enabled &PCIE_LINK_STATE_L1SS )) +pcie_config_aspm_l1ss (link,0 ); + +if (no_l1) { + link->aspm_support &= ~(PCIE_LINK_STATE_L1 |PCIE_LINK_STATE_L1SS ); + link->aspm_default &= ~(PCIE_LINK_STATE_L1 |PCIE_LINK_STATE_L1SS ); + link->aspm_enabled &= ~(PCIE_LINK_STATE_L1 |PCIE_LINK_STATE_L1SS ); + }else if (no_l1ss) { + link->aspm_support &= ~PCIE_LINK_STATE_L1SS ; + link->aspm_default &= ~PCIE_LINK_STATE_L1SS ; + link->aspm_enabled &= ~PCIE_LINK_STATE_L1SS ; + } +if (link->aspm_support &PCIE_LINK_STATE_L0S ) link->aspm_default |=PCIE_LINK_STATE_L0S ;if (link->aspm_support &PCIE_LINK_STATE_L1 )
@@ -926,9 +999,25 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)aspm_l1ss_init (link); -/* Restore L0s/L1 if they were enabled */ +/* Save default state */ + link->aspm_default = link->aspm_enabled; + +pcie_aspm_override_default_link_state (link); + +/* +* Restore L0s/L1 if they were enabled, but don't restore any +* state a Devicetree override just disabled in aspm_support above. +*/ if (FIELD_GET (PCI_EXP_LNKCTL_ASPMC , child_lnkctl) ||FIELD_GET (PCI_EXP_LNKCTL_ASPMC , parent_lnkctl)) { +if (!(link->aspm_support &PCIE_LINK_STATE_L0S )) { + child_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L0S ; + parent_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L0S ; + } +if (!(link->aspm_support &PCIE_LINK_STATE_L1 )) { + child_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L1 ; + parent_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L1 ; + }pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL ,PCI_EXP_LNKCTL_ASPMC , parent_lnkctl &PCI_EXP_LNKCTL_ASPMC );
@@ -938,11 +1027,6 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist) child_lnkctl & PCI_EXP_LNKCTL_ASPMC); } -/* Save default state */ - link->aspm_default = link->aspm_enabled; - -pcie_aspm_override_default_link_state (link); -/* Setup initial capable state. Will be updated later */ link->aspm_capable = link->aspm_support;
@@ -956,42 +1040,6 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist) } } -/* Configure the ASPM L1 substates. Caller must disable L1 first. */ -static void pcie_config_aspm_l1ss (struct pcie_link_state *link,u32 state) -{ -u32 val =0 ; -struct pci_dev *child = link->downstream, *parent = link->pdev; - -if (state &PCIE_LINK_STATE_L1_1 ) - val |=PCI_L1SS_CTL1_ASPM_L1_1 ; -if (state &PCIE_LINK_STATE_L1_2 ) - val |=PCI_L1SS_CTL1_ASPM_L1_2 ; -if (state &PCIE_LINK_STATE_L1_1_PCIPM ) - val |=PCI_L1SS_CTL1_PCIPM_L1_1 ; -if (state &PCIE_LINK_STATE_L1_2_PCIPM ) - val |=PCI_L1SS_CTL1_PCIPM_L1_2 ; - -/* -* PCIe r6.2, sec 5.5.4, rules for enabling L1 PM Substates: -* - Clear L1.x enable bits at child first, then at parent -* - Set L1.x enable bits at parent first, then at child -* - ASPM/PCIPM L1.2 must be disabled while programming timing -* parameters -*/ - -/* Disable all L1 substates */ -pci_clear_and_set_config_dword (child, child->l1ss +PCI_L1SS_CTL1 , -PCI_L1SS_CTL1_L1SS_MASK ,0 ); -pci_clear_and_set_config_dword (parent, parent->l1ss +PCI_L1SS_CTL1 , -PCI_L1SS_CTL1_L1SS_MASK ,0 ); - -/* Enable what we need to enable */ -pci_clear_and_set_config_dword (parent, parent->l1ss +PCI_L1SS_CTL1 , -PCI_L1SS_CTL1_L1SS_MASK , val); -pci_clear_and_set_config_dword (child, child->l1ss +PCI_L1SS_CTL1 , -PCI_L1SS_CTL1_L1SS_MASK , val); -} -static void pcie_config_aspm_dev (struct pci_dev *pdev,u32 val) { pcie_capability_clear_and_set_word(pdev, PCI_EXP_LNKCTL,
Signature · 1 line
2.34.1
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Hardware ASPM state becomes desynchronized from the software-tracked `link->aspm_enabled` state…
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Hardware ASPM state becomes desynchronized from the software-tracked link->aspm_enabled state on links where firmware forced ASPM L0s/L1 on despite missing hardware capabilities, breaking sysfs reporting. --
--- Patch [3]: [PATCH v6 3/3] PCI/ASPM: Mask ASPM states based on Devicetree properties --- commit d3ca900c30e202e8adf6acdf7daf494224ae2179
Author: Krishna Chaitanya Chundru <[email protected]>
PCI/ASPM: Mask ASPM states based on Devicetree properties
This commit introduces Devicetree property checks to selectively disable ASPM states based on hardware constraints. It moves the default state override evaluation earlier in capability initialization and ensures that states disabled by Devicetree are explicitly cleared from the hardware configuration during the L0s/L1 restore phase.
Link [1]: https://github.com/devicetree-org/dt-schema/pull/188
Link [2]: https://github.com/devicetree-org/dt-schema/pull/190
Quoted 4 lines
drivers/pci/pcie/aspm.c +0
-0
[ ... ]
Quoted 14 lines
drivers/pci/pcie/aspm.c +9
-1@@ -926,9 +999,25 @@ static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist)aspm_l1ss_init (link); -/* Restore L0s/L1 if they were enabled */ +/* Save default state */ + link->aspm_default = link->aspm_enabled; + +pcie_aspm_override_default_link_state (link); + +/* +* Restore L0s/L1 if they were enabled, but don't restore any +* state a Devicetree override just disabled in aspm_support above. +*/
drivers/pci/pcie/aspm.c static void pcie_aspm_cap_init(struct pcie_link_state *link, int blacklist) · line 1020
Quoted 10 lines
drivers/pci/pcie/aspm.c +8
-0if (FIELD_GET (PCI_EXP_LNKCTL_ASPMC , child_lnkctl) ||FIELD_GET (PCI_EXP_LNKCTL_ASPMC , parent_lnkctl)) { +if (!(link->aspm_support &PCIE_LINK_STATE_L0S )) { + child_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L0S ; + parent_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L0S ; + } +if (!(link->aspm_support &PCIE_LINK_STATE_L1 )) { + child_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L1 ; + parent_lnkctl &= ~PCI_EXP_LNKCTL_ASPM_L1 ; + }
[Severity: Medium]
If firmware forced ASPM L0s or L1 on despite missing hardware capabilities, this block will correctly strip those unsupported bits before writing back to hardware. However, does this leave the software state desynchronized?
Earlier during initialization, link->aspm_enabled captures the original firmware settings. Since this restore block now selectively clears unsupported bits from the hardware write but doesn't synchronize those changes back to link->aspm_enabled, won't sysfs attributes and pcie_aspm_enabled() incorrectly report these states as active?
Could this lead to incorrect sysfs reporting until pci_enable_device() eventually triggers a heal of the state via pcie_config_aspm_link()?
Quoted 3 lines
drivers/pci/pcie/aspm.c +0
-0pcie_capability_clear_and_set_word (parent,PCI_EXP_LNKCTL ,PCI_EXP_LNKCTL_ASPMC , parent_lnkctl &PCI_EXP_LNKCTL_ASPMC );
[ ... ]
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=3
Quoted 49 lines · Krishna Chaitanya Chundru
Some platforms require selectively disabling specific ASPM states on a given PCIe link to avoid link instability or functional failures caused by board-level connectivity constraints such as PCB routing, connectors, slots, or external cabling.
Devicetree supports disabling ASPM L0s, L1, and L1 PM Substates via the 'aspm-no-l0s', 'aspm-no-l1' [1], and 'aspm-no-l1ss' [2] properties. However, the ASPM driver does not currently honor these properties when initializing the default link state.
When firmware enables L1 PM Substates before the kernel takes over, masking
aspm_supportalone is insufficient to disable them in hardware.pcie_config_aspm_link()guards L1SS configuration behind a check onaspm_capable, which is derived fromaspm_support. Onceaspm_supportis masked,pcie_config_aspm_l1ss()is never called, leaving firmware-enabled L1SS substates active in hardware.Fix this by introducing
pcie_link_has_aspm_override()to check for DT override properties on either endpoint of the link. Inpcie_aspm_override_default_link_state(), use it to:- Mask
aspm_support,aspm_default, andaspm_enabledfor any disabled state, so software's view of the link stays in sync with what is actually programmed in hardware. Leavingaspm_enabledstale would makepcie_aspm_enabled()and the aspm sysfs attributes report a state as active even after it has been masked, and could cause pcie_config_aspm_link()'s "already in requested state" check to skip reprogramming hardware to match.
- Explicitly callpcie_config_aspm_l1ss(link, 0)before maskingaspm_supportwhen firmware has L1SS active and DT requests disabling L1 or L1SS, sincepcie_config_aspm_link()will no longer do so onceaspm_capableis derived from the maskedaspm_support.Move the
aspm_defaultinitialization andpcie_aspm_override_default_link_state()call inpcie_aspm_cap_init()to before the "Restore L0s/L1" block.pcie_aspm_cap_init()disables L1 in hardware prior toaspm_l1ss_init()and re-enables it only in the restore block. Callingpcie_config_aspm_l1ss()while L1 is already disabled satisfies its precondition ("Caller must disable L1 first"), whereas the previous placement after the restore violated it.Since the restore block writes back the parent_lnkctl/child_lnkctl snapshot taken from hardware before the DT override ran, mask the L0s and L1 enable bits out of that snapshot for any state the override has just disabled in
aspm_support. Otherwise the restore step would unconditionally reprogram the link back to firmware's original L0s/L1 configuration, defeating the Devicetree override it is meant to enforce.
Quoted 6 lines
Move
pcie_config_aspm_l1ss()earlier in the file so it can be called frompcie_aspm_override_default_link_state().Link [1]: https://github.com/devicetree-org/dt-schema/pull/188
Link [2]: https://github.com/devicetree-org/dt-schema/pull/190Signed-off-by Krishna Chaitanya Chundru <[email protected]>
- Mani
Signature · 1 line
மணிவண்ணன் சதாசிவம்
Bjorn Helgaas
AppliedQuoted 1 line · Krishna Chaitanya Chundru
Signed-off-by Krishna Chaitanya Chundru <[email protected]>
Applied to pci/aspm for v7.3, thank you!
Quoted 50 lines
---
Changes in v6:
- Disable/enable each function L0s/L1. (sashiko)
- Move frompcie_capability_write_wordtopcie_capability_clear_and_set_word(Sashiko)
- Link to v5: https://patch.msgid.link/[email protected]Changes in v5:
- update all the child devices when enabling the ASPM back (sashiko).
- Link to v4: https://patch.msgid.link/[email protected]Changes in v4:
- Clearlink->aspm_enabledalongside aspm_support/aspm_default when an ASPM state is masked, so it no longer goes stale relative to hardware (sashiko).
- Mask the LNKCTL restore snapshot against the post-override
aspm_supportso the restore step can no longer re-enable a state the Devicetree override just disabled. (sashiko)
- Link to v3: https://patch.msgid.link/[email protected]Changes in v3:
- Movepcie_aspm_override_default_link_state()call inpcie_aspm_cap_init()to before the "Restore L0s/L1" block.pcie_aspm_cap_init()disables L1 in hardware prior toaspm_l1ss_init()and re-enables it only in the restore block. Callingpcie_config_aspm_l1ss()while L1 is already disabled satisfies its precondition ("Caller must disable L1 first"), whereas the previous placement after the restore violated it (sashiko).
- Link to v2: https://patch.msgid.link/[email protected]Changes in v2:
- Disable L1ss when L1 is disabled as pointed by sashiko.
- Disable L1ss if bootloader enables them but we are disabling via devicetree pointed by sashiko.
- Link to v1: https://patch.msgid.link/[email protected]---
Krishna Chaitanya Chundru (3):
PCI/ASPM: Usepcie_capability_clear_and_set_word()for ASPM disable/restore PCI/ASPM: Disable/restore ASPM on every function for multi-function devices PCI/ASPM: Mask ASPM states based on Devicetree propertiesdrivers/pci/pcie/aspm.c | 149 +++++++++++++++++++++++++++++++++--------------- 1 file changed, 102 insertions(+), 47 deletions(-)---
base-commit:f5098b6bae761e346ebcd9da7f95622c04733cff
change-id: 20260429-aspm-da55a54279e4Best regards,
Signature · 1 line
Krishna Chaitanya Chundru <[email protected]>