February 2025 - Linux-stable-mirror

[PATCH v3 2/3] arm64: hugetlb: Fix huge_ptep_get_and_clear() for non-present ptes

by Ryan Roberts

arm64 supports multiple huge_pte sizes. Some of the sizes are covered by a single pte entry at a particular level (PMD_SIZE, PUD_SIZE), and some are covered by multiple ptes at a particular level (CONT_PTE_SIZE, CONT_PMD_SIZE). So the function has to figure out the size from the huge_pte pointer. This was previously done by walking the pgtable to determine the level and by using the PTE_CONT bit to determine the number of ptes at the level. But the PTE_CONT bit is only valid when the pte is present. For non-present pte values (e.g. markers, migration entries), the previous implementation was therefore erroneously determining the size. There is at least one known caller in core-mm, move_huge_pte(), which may call huge_ptep_get_and_clear() for a non-present pte. So we must be robust to this case. Additionally the "regular" ptep_get_and_clear() is robust to being called for non-present ptes so it makes sense to follow the behavior. Fix this by using the new sz parameter which is now provided to the function. Additionally when clearing each pte in a contig range, don't gather the access and dirty bits if the pte is not present. An alternative approach that would not require API changes would be to store the PTE_CONT bit in a spare bit in the swap entry pte for the non-present case. But it felt cleaner to follow other APIs' lead and just pass in the size. As an aside, PTE_CONT is bit 52, which corresponds to bit 40 in the swap entry offset field (layout of non-present pte). Since hugetlb is never swapped to disk, this field will only be populated for markers, which always set this bit to 0 and hwpoison swap entries, which set the offset field to a PFN; So it would only ever be 1 for a 52-bit PVA system where memory in that high half was poisoned (I think!). So in practice, this bit would almost always be zero for non-present ptes and we would only clear the first entry if it was actually a contiguous block. That's probably a less severe symptom than if it was always interpreted as 1 and cleared out potentially-present neighboring PTEs. Cc: stable(a)vger.kernel.org Fixes: 66b3923a1a0f ("arm64: hugetlb: add support for PTE contiguous bit") Reviewed-by: Catalin Marinas <catalin.marinas(a)arm.com> Signed-off-by: Ryan Roberts <ryan.roberts(a)arm.com> tmp --- arch/arm64/mm/hugetlbpage.c | 53 ++++++++++++++----------------------- 1 file changed, 20 insertions(+), 33 deletions(-) diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c index 06db4649af91..b3a7fafe8892 100644 --- a/arch/arm64/mm/hugetlbpage.c +++ b/arch/arm64/mm/hugetlbpage.c @@ -100,20 +100,11 @@ static int find_num_contig(struct mm_struct *mm, unsigned long addr, static inline int num_contig_ptes(unsigned long size, size_t *pgsize) { - int contig_ptes = 0; + int contig_ptes = 1; *pgsize = size; switch (size) { -#ifndef __PAGETABLE_PMD_FOLDED - case PUD_SIZE: - if (pud_sect_supported()) - contig_ptes = 1; - break; -#endif - case PMD_SIZE: - contig_ptes = 1; - break; case CONT_PMD_SIZE: *pgsize = PMD_SIZE; contig_ptes = CONT_PMDS; @@ -122,6 +113,8 @@ static inline int num_contig_ptes(unsigned long size, size_t *pgsize) *pgsize = PAGE_SIZE; contig_ptes = CONT_PTES; break; + default: + WARN_ON(!__hugetlb_valid_size(size)); } return contig_ptes; @@ -163,24 +156,23 @@ static pte_t get_clear_contig(struct mm_struct *mm, unsigned long pgsize, unsigned long ncontig) { - pte_t orig_pte = __ptep_get(ptep); - unsigned long i; - - for (i = 0; i < ncontig; i++, addr += pgsize, ptep++) { - pte_t pte = __ptep_get_and_clear(mm, addr, ptep); - - /* - * If HW_AFDBM is enabled, then the HW could turn on - * the dirty or accessed bit for any page in the set, - * so check them all. - */ - if (pte_dirty(pte)) - orig_pte = pte_mkdirty(orig_pte); - - if (pte_young(pte)) - orig_pte = pte_mkyoung(orig_pte); + pte_t pte, tmp_pte; + bool present; + + pte = __ptep_get_and_clear(mm, addr, ptep); + present = pte_present(pte); + while (--ncontig) { + ptep++; + addr += pgsize; + tmp_pte = __ptep_get_and_clear(mm, addr, ptep); + if (present) { + if (pte_dirty(tmp_pte)) + pte = pte_mkdirty(pte); + if (pte_young(tmp_pte)) + pte = pte_mkyoung(pte); + } } - return orig_pte; + return pte; } static pte_t get_clear_contig_flush(struct mm_struct *mm, @@ -401,13 +393,8 @@ pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, { int ncontig; size_t pgsize; - pte_t orig_pte = __ptep_get(ptep); - - if (!pte_cont(orig_pte)) - return __ptep_get_and_clear(mm, addr, ptep); - - ncontig = find_num_contig(mm, addr, ptep, &pgsize); + ncontig = num_contig_ptes(sz, &pgsize); return get_clear_contig(mm, addr, ptep, pgsize, ncontig); } -- 2.43.0

5 days, 2 hours

2
2
0 0

[PATCH 8/8] arm64: dts: qcom: x1e78100-t14s: fix missing HID supplies

by Johan Hovold

Add the missing HID supplies to avoid relying on other consumers to keep them on. This also avoids the following warnings on boot: i2c_hid_of 0-0010: supply vdd not found, using dummy regulator i2c_hid_of 0-0010: supply vddl not found, using dummy regulator i2c_hid_of 1-0015: supply vdd not found, using dummy regulator i2c_hid_of 1-002c: supply vdd not found, using dummy regulator i2c_hid_of 1-0015: supply vddl not found, using dummy regulator i2c_hid_of 1-002c: supply vddl not found, using dummy regulator i2c_hid_of 1-003a: supply vdd not found, using dummy regulator i2c_hid_of 1-003a: supply vddl not found, using dummy regulator Note that VCC3B is also used for things like the modem which are not yet described so mark the regulator as always-on for now. Fixes: 7d1cbe2f4985 ("arm64: dts: qcom: Add X1E78100 ThinkPad T14s Gen 6") Cc: stable(a)vger.kernel.org # 6.12 Cc: Konrad Dybcio <konradybcio(a)kernel.org> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- .../qcom/x1e78100-lenovo-thinkpad-t14s.dts | 43 +++++++++++++++++++ 1 file changed, 43 insertions(+) diff --git a/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts b/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts index 7f756ce48d2f..3ff0c65e374c 100644 --- a/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts +++ b/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts @@ -9,6 +9,7 @@ #include <dt-bindings/gpio/gpio.h> #include <dt-bindings/input/gpio-keys.h> #include <dt-bindings/input/input.h> +#include <dt-bindings/pinctrl/qcom,pmic-gpio.h> #include <dt-bindings/regulator/qcom,rpmh-regulator.h> #include "x1e80100.dtsi" @@ -153,6 +154,23 @@ vreg_edp_3p3: regulator-edp-3p3 { regulator-boot-on; }; + vreg_misc_3p3: regulator-misc-3p3 { + compatible = "regulator-fixed"; + + regulator-name = "VCC3B"; + regulator-min-microvolt = <3300000>; + regulator-max-microvolt = <3300000>; + + gpio = <&pm8550ve_8_gpios 6 GPIO_ACTIVE_HIGH>; + enable-active-high; + + pinctrl-0 = <&misc_3p3_reg_en>; + pinctrl-names = "default"; + + regulator-boot-on; + regulator-always-on; + }; + vreg_nvme: regulator-nvme { compatible = "regulator-fixed"; @@ -580,6 +598,9 @@ touchpad@15 { hid-descr-addr = <0x1>; interrupts-extended = <&tlmm 3 IRQ_TYPE_LEVEL_LOW>; + vdd-supply = <&vreg_misc_3p3>; + vddl-supply = <&vreg_l12b_1p2>; + wakeup-source; }; @@ -591,6 +612,9 @@ touchpad@2c { hid-descr-addr = <0x20>; interrupts-extended = <&tlmm 3 IRQ_TYPE_LEVEL_LOW>; + vdd-supply = <&vreg_misc_3p3>; + vddl-supply = <&vreg_l12b_1p2>; + wakeup-source; }; @@ -602,6 +626,9 @@ keyboard@3a { hid-descr-addr = <0x1>; interrupts-extended = <&tlmm 67 IRQ_TYPE_LEVEL_LOW>; + vdd-supply = <&vreg_misc_3p3>; + vddl-supply = <&vreg_l15b_1p8>; + pinctrl-0 = <&kybd_default>; pinctrl-names = "default"; @@ -670,6 +697,9 @@ touchscreen@10 { hid-descr-addr = <0x1>; interrupts-extended = <&tlmm 51 IRQ_TYPE_LEVEL_LOW>; + vdd-supply = <&vreg_misc_3p3>; + vddl-supply = <&vreg_l15b_1p8>; + pinctrl-0 = <&ts0_default>; pinctrl-names = "default"; }; @@ -779,6 +809,19 @@ &pcie6a_phy { status = "okay"; }; +&pm8550ve_8_gpios { + misc_3p3_reg_en: misc-3p3-reg-en-state { + pins = "gpio6"; + function = "normal"; + bias-disable; + drive-push-pull; + input-disable; + output-enable; + power-source = <1>; /* 1.8 V */ + qcom,drive-strength = <PMIC_GPIO_STRENGTH_LOW>; + }; +}; + &pmc8380_3_gpios { edp_bl_en: edp-bl-en-state { pins = "gpio4"; -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 7/8] arm64: dts: qcom: x1e80100-qcp: mark l12b and l15b always-on

by Johan Hovold

The l12b and l15b supplies are used by components that are not (fully) described (and some never will be) and must never be disabled. Mark the regulators as always-on to prevent them from being disabled, for example, when consumers probe defer or suspend. Fixes: af16b00578a7 ("arm64: dts: qcom: Add base X1E80100 dtsi and the QCP dts") Cc: stable(a)vger.kernel.org # 6.8 Cc: Rajendra Nayak <quic_rjendra(a)quicinc.com> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- arch/arm64/boot/dts/qcom/x1e80100-qcp.dts | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm64/boot/dts/qcom/x1e80100-qcp.dts b/arch/arm64/boot/dts/qcom/x1e80100-qcp.dts index ec594628304a..8f366bf61bbd 100644 --- a/arch/arm64/boot/dts/qcom/x1e80100-qcp.dts +++ b/arch/arm64/boot/dts/qcom/x1e80100-qcp.dts @@ -437,6 +437,7 @@ vreg_l12b_1p2: ldo12 { regulator-min-microvolt = <1200000>; regulator-max-microvolt = <1200000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l13b_3p0: ldo13 { @@ -458,6 +459,7 @@ vreg_l15b_1p8: ldo15 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l16b_2p9: ldo16 { -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 6/8] arm64: dts: qcom: x1e80100-yoga-slim7x: mark l12b and l15b always-on

by Johan Hovold

The l12b and l15b supplies are used by components that are not (fully) described (and some never will be) and must never be disabled. Mark the regulators as always-on to prevent them from being disabled, for example, when consumers probe defer or suspend. Fixes: 45247fe17db2 ("arm64: dts: qcom: x1e80100: add Lenovo Thinkpad Yoga slim 7x devicetree") Cc: stable(a)vger.kernel.org # 6.11 Cc: Srinivas Kandagatla <srinivas.kandagatla(a)linaro.org> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- arch/arm64/boot/dts/qcom/x1e80100-lenovo-yoga-slim7x.dts | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/arm64/boot/dts/qcom/x1e80100-lenovo-yoga-slim7x.dts b/arch/arm64/boot/dts/qcom/x1e80100-lenovo-yoga-slim7x.dts index a3d53f2ba2c3..9d4ba9728355 100644 --- a/arch/arm64/boot/dts/qcom/x1e80100-lenovo-yoga-slim7x.dts +++ b/arch/arm64/boot/dts/qcom/x1e80100-lenovo-yoga-slim7x.dts @@ -290,6 +290,7 @@ vreg_l12b_1p2: ldo12 { regulator-min-microvolt = <1200000>; regulator-max-microvolt = <1200000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l14b_3p0: ldo14 { @@ -304,8 +305,8 @@ vreg_l15b_1p8: ldo15 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; - }; regulators-1 { -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 5/8] arm64: dts: qcom: x1e80100-hp-x14: mark l12b and l15b always-on

by Johan Hovold

The l12b and l15b supplies are used by components that are not (fully) described (and some never will be) and must never be disabled. Mark the regulators as always-on to prevent them from being disabled, for example, when consumers probe defer or suspend. Fixes: 6f18b8d4142c ("arm64: dts: qcom: x1e80100-hp-x14: dt for HP Omnibook X Laptop 14") Cc: stable(a)vger.kernel.org # 6.14 Cc: Jens Glathe <jens.glathe(a)oldschoolsolutions.biz> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- arch/arm64/boot/dts/qcom/x1e80100-hp-omnibook-x14.dts | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm64/boot/dts/qcom/x1e80100-hp-omnibook-x14.dts b/arch/arm64/boot/dts/qcom/x1e80100-hp-omnibook-x14.dts index cd860a246c45..ab5addb33b7a 100644 --- a/arch/arm64/boot/dts/qcom/x1e80100-hp-omnibook-x14.dts +++ b/arch/arm64/boot/dts/qcom/x1e80100-hp-omnibook-x14.dts @@ -633,6 +633,7 @@ vreg_l12b_1p2: ldo12 { regulator-min-microvolt = <1200000>; regulator-max-microvolt = <1200000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l13b_3p0: ldo13 { @@ -654,6 +655,7 @@ vreg_l15b_1p8: ldo15 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l16b_2p9: ldo16 { -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 3/8] arm64: dts: qcom: x1e001de-devkit: mark l12b and l15b always-on

by Johan Hovold

The l12b and l15b supplies are used by components that are not (fully) described (and some never will be) and must never be disabled. Mark the regulators as always-on to prevent them from being disabled, for example, when consumers probe defer or suspend. Fixes: 7b8a31e82b87 ("arm64: dts: qcom: Add X1E001DE Snapdragon Devkit for Windows") Cc: stable(a)vger.kernel.org # 6.14 Cc: Sibi Sankar <quic_sibis(a)quicinc.com> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- arch/arm64/boot/dts/qcom/x1e001de-devkit.dts | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm64/boot/dts/qcom/x1e001de-devkit.dts b/arch/arm64/boot/dts/qcom/x1e001de-devkit.dts index 5e3970b26e2f..f92bda2d34f2 100644 --- a/arch/arm64/boot/dts/qcom/x1e001de-devkit.dts +++ b/arch/arm64/boot/dts/qcom/x1e001de-devkit.dts @@ -507,6 +507,7 @@ vreg_l12b_1p2: ldo12 { regulator-min-microvolt = <1200000>; regulator-max-microvolt = <1200000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l13b_3p0: ldo13 { @@ -528,6 +529,7 @@ vreg_l15b_1p8: ldo15 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l16b_2p9: ldo16 { -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 2/8] arm64: dts: qcom: x1e78100-t14s: mark l12b and l15b always-on

by Johan Hovold

The l12b and l15b supplies are used by components that are not (fully) described (and some never will be) and must never be disabled. Mark the regulators as always-on to prevent them from being disabled, for example, when consumers probe defer or suspend. Fixes: 7d1cbe2f4985 ("arm64: dts: qcom: Add X1E78100 ThinkPad T14s Gen 6") Cc: stable(a)vger.kernel.org # 6.12 Cc: Konrad Dybcio <konradybcio(a)kernel.org> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts b/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts index b2c2347f54fa..7f756ce48d2f 100644 --- a/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts +++ b/arch/arm64/boot/dts/qcom/x1e78100-lenovo-thinkpad-t14s.dts @@ -344,6 +344,7 @@ vreg_l12b_1p2: ldo12 { regulator-min-microvolt = <1200000>; regulator-max-microvolt = <1200000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l13b_3p0: ldo13 { @@ -365,6 +366,7 @@ vreg_l15b_1p8: ldo15 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l17b_2p5: ldo17 { -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 1/8] arm64: dts: qcom: x1e80100-crd: mark l12b and l15b always-on

by Johan Hovold

The l12b and l15b supplies are used by components that are not (fully) described (and some never will be) and must never be disabled. Mark the regulators as always-on to prevent them from being disabled, for example, when consumers probe defer or suspend. Fixes: bd50b1f5b6f3 ("arm64: dts: qcom: x1e80100: Add Compute Reference Device") Cc: stable(a)vger.kernel.org # 6.8 Cc: Abel Vesa <abel.vesa(a)linaro.org> Cc: Rajendra Nayak <quic_rjendra(a)quicinc.com> Cc: Sibi Sankar <quic_sibis(a)quicinc.com> Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- arch/arm64/boot/dts/qcom/x1e80100-crd.dts | 2 ++ 1 file changed, 2 insertions(+) diff --git a/arch/arm64/boot/dts/qcom/x1e80100-crd.dts b/arch/arm64/boot/dts/qcom/x1e80100-crd.dts index ff5b3472fafd..ffce8f1eb2e1 100644 --- a/arch/arm64/boot/dts/qcom/x1e80100-crd.dts +++ b/arch/arm64/boot/dts/qcom/x1e80100-crd.dts @@ -437,6 +437,7 @@ vreg_l12b_1p2: ldo12 { regulator-min-microvolt = <1200000>; regulator-max-microvolt = <1200000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l13b_3p0: ldo13 { @@ -458,6 +459,7 @@ vreg_l15b_1p8: ldo15 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>; + regulator-always-on; }; vreg_l16b_2p9: ldo16 { -- 2.45.3

5 days, 3 hours

1
0
0 0

[PATCH 0/6] drm/v3d: Fix GPU reset issues on the Raspberry Pi 5

by Maíra Canal

This series addresses GPU reset issues reported in [1], where running a long compute job would trigger repeated GPU resets, leading to a UI freeze. Patches #1 and #2 prevent the same faulty job from being resubmitted in a loop, mitigating the first cause of the issue. However, the issue isn't entirely solved. Even with only a single GPU reset, the UI still freezes on the Raspberry Pi 5, indicating a GPU hang. Patches #3 to #5 address this by properly configuring the V3D_SMS registers, which are required for power management and resets in V3D 7.1. Patch #6 updates the DT maintainership, replacing Emma with the current v3d driver maintainer. [1] https://github.com/raspberrypi/linux/issues/6660 Best Regards, - Maíra --- Maíra Canal (6): drm/v3d: Don't run jobs that have errors flagged in its fence drm/v3d: Set job pointer to NULL when the job's fence has an error drm/v3d: Associate a V3D tech revision to all supported devices dt-bindings: gpu: v3d: Add SMS to the registers' list drm/v3d: Use V3D_SMS registers for power on/off and reset on V3D 7.x dt-bindings: gpu: Add V3D driver maintainer as DT maintainer .../devicetree/bindings/gpu/brcm,bcm-v3d.yaml | 8 +-- drivers/gpu/drm/v3d/v3d_drv.c | 58 ++++++++++++++++++++-- drivers/gpu/drm/v3d/v3d_drv.h | 18 +++++++ drivers/gpu/drm/v3d/v3d_gem.c | 17 +++++++ drivers/gpu/drm/v3d/v3d_regs.h | 26 ++++++++++ drivers/gpu/drm/v3d/v3d_sched.c | 23 +++++++-- 6 files changed, 140 insertions(+), 10 deletions(-) --- base-commit: 099b79f94366f3110783301e20d8136d762247f8 change-id: 20250224-v3d-gpu-reset-fixes-2d21fc70711d

5 days, 3 hours

2
2
0 0

[PATCH 3/4] drm/xe: Fix fault mode invalidation with unbind

by Thomas Hellström

Fix fault mode invalidation racing with unbind leading to the PTE zapping potentially traversing an invalid page-table tree. Do this by holding the notifier lock across PTE zapping. This might transfer any contention waiting on the notifier seqlock read side to the notifier lock read side, but that shouldn't be a major problem. At the same time get rid of the open-coded invalidation in the bind code by relying on the notifier even when the vma bind is not yet committed. Finally let userptr invalidation call a dedicated xe_vm function performing a full invalidation. Fixes: e8babb280b5e ("drm/xe: Convert multiple bind ops into single job") Cc: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: Matthew Auld <matthew.auld(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.12+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_pt.c | 38 ++++------------ drivers/gpu/drm/xe/xe_vm.c | 78 ++++++++++++++++++++------------ drivers/gpu/drm/xe/xe_vm.h | 8 ++++ drivers/gpu/drm/xe/xe_vm_types.h | 4 +- 4 files changed, 68 insertions(+), 60 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c index 1ddcc7e79a93..12a627a23eb4 100644 --- a/drivers/gpu/drm/xe/xe_pt.c +++ b/drivers/gpu/drm/xe/xe_pt.c @@ -1213,42 +1213,22 @@ static int vma_check_userptr(struct xe_vm *vm, struct xe_vma *vma, return 0; uvma = to_userptr_vma(vma); - notifier_seq = uvma->userptr.notifier_seq; + if (xe_pt_userptr_inject_eagain(uvma)) + xe_vma_userptr_force_invalidate(uvma); - if (uvma->userptr.initial_bind && !xe_vm_in_fault_mode(vm)) - return 0; + notifier_seq = uvma->userptr.notifier_seq; if (!mmu_interval_read_retry(&uvma->userptr.notifier, - notifier_seq) && - !xe_pt_userptr_inject_eagain(uvma)) + notifier_seq)) return 0; - if (xe_vm_in_fault_mode(vm)) { + if (xe_vm_in_fault_mode(vm)) return -EAGAIN; - } else { - spin_lock(&vm->userptr.invalidated_lock); - list_move_tail(&uvma->userptr.invalidate_link, - &vm->userptr.invalidated); - spin_unlock(&vm->userptr.invalidated_lock); - - if (xe_vm_in_preempt_fence_mode(vm)) { - struct dma_resv_iter cursor; - struct dma_fence *fence; - long err; - - dma_resv_iter_begin(&cursor, xe_vm_resv(vm), - DMA_RESV_USAGE_BOOKKEEP); - dma_resv_for_each_fence_unlocked(&cursor, fence) - dma_fence_enable_sw_signaling(fence); - dma_resv_iter_end(&cursor); - - err = dma_resv_wait_timeout(xe_vm_resv(vm), - DMA_RESV_USAGE_BOOKKEEP, - false, MAX_SCHEDULE_TIMEOUT); - XE_WARN_ON(err <= 0); - } - } + /* + * Just continue the operation since exec or rebind worker + * will take care of rebinding. + */ return 0; } diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 4c1ca47667ad..37d773c0b729 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -580,51 +580,26 @@ static void preempt_rebind_work_func(struct work_struct *w) trace_xe_vm_rebind_worker_exit(vm); } -static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni, - const struct mmu_notifier_range *range, - unsigned long cur_seq) +static void __vma_userptr_invalidate(struct xe_vm *vm, struct xe_userptr_vma *uvma) { - struct xe_userptr *userptr = container_of(mni, typeof(*userptr), notifier); - struct xe_userptr_vma *uvma = container_of(userptr, typeof(*uvma), userptr); + struct xe_userptr *userptr = &uvma->userptr; struct xe_vma *vma = &uvma->vma; - struct xe_vm *vm = xe_vma_vm(vma); struct dma_resv_iter cursor; struct dma_fence *fence; long err; - xe_assert(vm->xe, xe_vma_is_userptr(vma)); - trace_xe_vma_userptr_invalidate(vma); - - if (!mmu_notifier_range_blockable(range)) - return false; - - vm_dbg(&xe_vma_vm(vma)->xe->drm, - "NOTIFIER: addr=0x%016llx, range=0x%016llx", - xe_vma_start(vma), xe_vma_size(vma)); - - down_write(&vm->userptr.notifier_lock); - mmu_interval_set_seq(mni, cur_seq); - - /* No need to stop gpu access if the userptr is not yet bound. */ - if (!userptr->initial_bind) { - up_write(&vm->userptr.notifier_lock); - return true; - } - /* * Tell exec and rebind worker they need to repin and rebind this * userptr. */ if (!xe_vm_in_fault_mode(vm) && - !(vma->gpuva.flags & XE_VMA_DESTROYED) && vma->tile_present) { + !(vma->gpuva.flags & XE_VMA_DESTROYED)) { spin_lock(&vm->userptr.invalidated_lock); list_move_tail(&userptr->invalidate_link, &vm->userptr.invalidated); spin_unlock(&vm->userptr.invalidated_lock); } - up_write(&vm->userptr.notifier_lock); - /* * Preempt fences turn into schedule disables, pipeline these. * Note that even in fault mode, we need to wait for binds and @@ -642,11 +617,35 @@ static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni, false, MAX_SCHEDULE_TIMEOUT); XE_WARN_ON(err <= 0); - if (xe_vm_in_fault_mode(vm)) { + if (xe_vm_in_fault_mode(vm) && userptr->initial_bind) { err = xe_vm_invalidate_vma(vma); XE_WARN_ON(err); } +} + +static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni, + const struct mmu_notifier_range *range, + unsigned long cur_seq) +{ + struct xe_userptr_vma *uvma = container_of(mni, typeof(*uvma), userptr.notifier); + struct xe_vma *vma = &uvma->vma; + struct xe_vm *vm = xe_vma_vm(vma); + + xe_assert(vm->xe, xe_vma_is_userptr(vma)); + trace_xe_vma_userptr_invalidate(vma); + + if (!mmu_notifier_range_blockable(range)) + return false; + vm_dbg(&xe_vma_vm(vma)->xe->drm, + "NOTIFIER: addr=0x%016llx, range=0x%016llx", + xe_vma_start(vma), xe_vma_size(vma)); + + down_write(&vm->userptr.notifier_lock); + mmu_interval_set_seq(mni, cur_seq); + + __vma_userptr_invalidate(vm, uvma); + up_write(&vm->userptr.notifier_lock); trace_xe_vma_userptr_invalidate_complete(vma); return true; @@ -656,6 +655,27 @@ static const struct mmu_interval_notifier_ops vma_userptr_notifier_ops = { .invalidate = vma_userptr_invalidate, }; +#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT) +/** + * xe_vma_userptr_force_invalidate() - force invalidate a userptr + * @uvma: The userptr vma to invalidate + * + * Perform a forced userptr invalidation for testing purposes. + */ +void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma) +{ + struct xe_vm *vm = xe_vma_vm(&uvma->vma); + + lockdep_assert_held_write(&vm->lock); + lockdep_assert_held(&vm->userptr.notifier_lock); + + if (!mmu_interval_read_retry(&uvma->userptr.notifier, + uvma->userptr.notifier_seq)) + uvma->userptr.notifier_seq -= 2; + __vma_userptr_invalidate(vm, uvma); +} +#endif + int xe_vm_userptr_pin(struct xe_vm *vm) { struct xe_userptr_vma *uvma, *next; diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h index 7c8e39049223..f5d835271350 100644 --- a/drivers/gpu/drm/xe/xe_vm.h +++ b/drivers/gpu/drm/xe/xe_vm.h @@ -287,4 +287,12 @@ struct xe_vm_snapshot *xe_vm_snapshot_capture(struct xe_vm *vm); void xe_vm_snapshot_capture_delayed(struct xe_vm_snapshot *snap); void xe_vm_snapshot_print(struct xe_vm_snapshot *snap, struct drm_printer *p); void xe_vm_snapshot_free(struct xe_vm_snapshot *snap); + +#if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT) +void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma); +#else +static inline void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma) +{ +} +#endif #endif diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index 52467b9b5348..1fe79bf23b6b 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -228,8 +228,8 @@ struct xe_vm { * up for revalidation. Protected from access with the * @invalidated_lock. Removing items from the list * additionally requires @lock in write mode, and adding - * items to the list requires the @userptr.notifer_lock in - * write mode. + * items to the list requires either the @userptr.notifer_lock in + * write mode, OR @lock in write mode. */ struct list_head invalidated; } userptr; -- 2.48.1

5 days, 4 hours

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror February 2025