| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Thu, 27 Jan 2022 19:09:36 +0100
From: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To: linux-kernel@...r.kernel.org
Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
stable@...r.kernel.org, Tvrtko Ursulin <tvrtko.ursulin@...el.com>,
Sushma Venkatesh Reddy <sushma.venkatesh.reddy@...el.com>,
Daniel Vetter <daniel.vetter@...ll.ch>,
Dave Airlie <airlied@...hat.com>,
Jon Bloomfield <jon.bloomfield@...el.com>,
Joonas Lahtinen <joonas.lahtinen@...ux.intel.com>,
Jani Nikula <jani.nikula@...el.com>,
Linus Torvalds <torvalds@...ux-foundation.org>
Subject: [PATCH 5.16 1/9] drm/i915: Flush TLBs before releasing backing store
From: Tvrtko Ursulin <tvrtko.ursulin@...el.com>
commit 7938d61591d33394a21bdd7797a245b65428f44c upstream.
We need to flush TLBs before releasing backing store otherwise userspace
is able to encounter stale entries if a) it is not declaring access to
certain buffers and b) it races with the backing store release from a
such undeclared execution already executing on the GPU in parallel.
The approach taken is to mark any buffer objects which were ever bound
to the GPU and to trigger a serialized TLB flush when their backing
store is released.
Alternatively the flushing could be done on VMA unbind, at which point
we would be able to ascertain whether there is potential a parallel GPU
execution (which could race), but essentially it boils down to paying
the cost of TLB flushes potentially needlessly at VMA unbind time (when
the backing store is not known to be going away so not needed for
safety), versus potentially needlessly at backing store relase time
(since we at that point cannot tell whether there is anything executing
on the GPU which uses that object).
Thereforce simplicity of implementation has been chosen for now with
scope to benchmark and refine later as required.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@...el.com>
Reported-by: Sushma Venkatesh Reddy <sushma.venkatesh.reddy@...el.com>
Reviewed-by: Daniel Vetter <daniel.vetter@...ll.ch>
Acked-by: Dave Airlie <airlied@...hat.com>
Cc: Daniel Vetter <daniel.vetter@...ll.ch>
Cc: Jon Bloomfield <jon.bloomfield@...el.com>
Cc: Joonas Lahtinen <joonas.lahtinen@...ux.intel.com>
Cc: Jani Nikula <jani.nikula@...el.com>
Cc: stable@...r.kernel.org
Signed-off-by: Linus Torvalds <torvalds@...ux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
---
drivers/gpu/drm/i915/gem/i915_gem_object_types.h | 1
drivers/gpu/drm/i915/gem/i915_gem_pages.c | 10 ++
drivers/gpu/drm/i915/gt/intel_gt.c | 102 +++++++++++++++++++++++
drivers/gpu/drm/i915/gt/intel_gt.h | 2
drivers/gpu/drm/i915/gt/intel_gt_types.h | 2
drivers/gpu/drm/i915/i915_reg.h | 11 ++
drivers/gpu/drm/i915/i915_vma.c | 3
drivers/gpu/drm/i915/intel_uncore.c | 26 ++++-
drivers/gpu/drm/i915/intel_uncore.h | 2
9 files changed, 155 insertions(+), 4 deletions(-)
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -305,6 +305,7 @@ struct drm_i915_gem_object {
#define I915_BO_READONLY BIT(6)
#define I915_TILING_QUIRK_BIT 7 /* unknown swizzling; do not release! */
#define I915_BO_PROTECTED BIT(8)
+#define I915_BO_WAS_BOUND_BIT 9
/**
* @mem_flags - Mutable placement-related flags
*
--- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
@@ -10,6 +10,8 @@
#include "i915_gem_lmem.h"
#include "i915_gem_mman.h"
+#include "gt/intel_gt.h"
+
void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
struct sg_table *pages,
unsigned int sg_page_sizes)
@@ -217,6 +219,14 @@ __i915_gem_object_unset_pages(struct drm
__i915_gem_object_reset_page_iter(obj);
obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;
+ if (test_and_clear_bit(I915_BO_WAS_BOUND_BIT, &obj->flags)) {
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ intel_wakeref_t wakeref;
+
+ with_intel_runtime_pm_if_active(&i915->runtime_pm, wakeref)
+ intel_gt_invalidate_tlbs(&i915->gt);
+ }
+
return pages;
}
--- a/drivers/gpu/drm/i915/gt/intel_gt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt.c
@@ -30,6 +30,8 @@ void intel_gt_init_early(struct intel_gt
spin_lock_init(>->irq_lock);
+ mutex_init(>->tlb_invalidate_lock);
+
INIT_LIST_HEAD(>->closed_vma);
spin_lock_init(>->closed_lock);
@@ -907,3 +909,103 @@ void intel_gt_info_print(const struct in
intel_sseu_dump(&info->sseu, p);
}
+
+struct reg_and_bit {
+ i915_reg_t reg;
+ u32 bit;
+};
+
+static struct reg_and_bit
+get_reg_and_bit(const struct intel_engine_cs *engine, const bool gen8,
+ const i915_reg_t *regs, const unsigned int num)
+{
+ const unsigned int class = engine->class;
+ struct reg_and_bit rb = { };
+
+ if (drm_WARN_ON_ONCE(&engine->i915->drm,
+ class >= num || !regs[class].reg))
+ return rb;
+
+ rb.reg = regs[class];
+ if (gen8 && class == VIDEO_DECODE_CLASS)
+ rb.reg.reg += 4 * engine->instance; /* GEN8_M2TCR */
+ else
+ rb.bit = engine->instance;
+
+ rb.bit = BIT(rb.bit);
+
+ return rb;
+}
+
+void intel_gt_invalidate_tlbs(struct intel_gt *gt)
+{
+ static const i915_reg_t gen8_regs[] = {
+ [RENDER_CLASS] = GEN8_RTCR,
+ [VIDEO_DECODE_CLASS] = GEN8_M1TCR, /* , GEN8_M2TCR */
+ [VIDEO_ENHANCEMENT_CLASS] = GEN8_VTCR,
+ [COPY_ENGINE_CLASS] = GEN8_BTCR,
+ };
+ static const i915_reg_t gen12_regs[] = {
+ [RENDER_CLASS] = GEN12_GFX_TLB_INV_CR,
+ [VIDEO_DECODE_CLASS] = GEN12_VD_TLB_INV_CR,
+ [VIDEO_ENHANCEMENT_CLASS] = GEN12_VE_TLB_INV_CR,
+ [COPY_ENGINE_CLASS] = GEN12_BLT_TLB_INV_CR,
+ };
+ struct drm_i915_private *i915 = gt->i915;
+ struct intel_uncore *uncore = gt->uncore;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+ const i915_reg_t *regs;
+ unsigned int num = 0;
+
+ if (I915_SELFTEST_ONLY(gt->awake == -ENODEV))
+ return;
+
+ if (GRAPHICS_VER(i915) == 12) {
+ regs = gen12_regs;
+ num = ARRAY_SIZE(gen12_regs);
+ } else if (GRAPHICS_VER(i915) >= 8 && GRAPHICS_VER(i915) <= 11) {
+ regs = gen8_regs;
+ num = ARRAY_SIZE(gen8_regs);
+ } else if (GRAPHICS_VER(i915) < 8) {
+ return;
+ }
+
+ if (drm_WARN_ONCE(&i915->drm, !num,
+ "Platform does not implement TLB invalidation!"))
+ return;
+
+ GEM_TRACE("\n");
+
+ assert_rpm_wakelock_held(&i915->runtime_pm);
+
+ mutex_lock(>->tlb_invalidate_lock);
+ intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL);
+
+ for_each_engine(engine, gt, id) {
+ /*
+ * HW architecture suggest typical invalidation time at 40us,
+ * with pessimistic cases up to 100us and a recommendation to
+ * cap at 1ms. We go a bit higher just in case.
+ */
+ const unsigned int timeout_us = 100;
+ const unsigned int timeout_ms = 4;
+ struct reg_and_bit rb;
+
+ rb = get_reg_and_bit(engine, regs == gen8_regs, regs, num);
+ if (!i915_mmio_reg_offset(rb.reg))
+ continue;
+
+ intel_uncore_write_fw(uncore, rb.reg, rb.bit);
+ if (__intel_wait_for_register_fw(uncore,
+ rb.reg, rb.bit, 0,
+ timeout_us, timeout_ms,
+ NULL))
+ drm_err_ratelimited(>->i915->drm,
+ "%s TLB invalidation did not complete in %ums!\n",
+ engine->name, timeout_ms);
+ }
+
+ intel_uncore_forcewake_put_delayed(uncore, FORCEWAKE_ALL);
+ mutex_unlock(>->tlb_invalidate_lock);
+}
--- a/drivers/gpu/drm/i915/gt/intel_gt.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt.h
@@ -90,4 +90,6 @@ void intel_gt_info_print(const struct in
void intel_gt_watchdog_work(struct work_struct *work);
+void intel_gt_invalidate_tlbs(struct intel_gt *gt);
+
#endif /* __INTEL_GT_H__ */
--- a/drivers/gpu/drm/i915/gt/intel_gt_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h
@@ -73,6 +73,8 @@ struct intel_gt {
struct intel_uc uc;
+ struct mutex tlb_invalidate_lock;
+
struct i915_wa_list wa_list;
struct intel_gt_timelines {
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -2697,6 +2697,12 @@ static inline bool i915_mmio_reg_valid(i
#define GAMT_CHKN_DISABLE_DYNAMIC_CREDIT_SHARING (1 << 28)
#define GAMT_CHKN_DISABLE_I2M_CYCLE_ON_WR_PORT (1 << 24)
+#define GEN8_RTCR _MMIO(0x4260)
+#define GEN8_M1TCR _MMIO(0x4264)
+#define GEN8_M2TCR _MMIO(0x4268)
+#define GEN8_BTCR _MMIO(0x426c)
+#define GEN8_VTCR _MMIO(0x4270)
+
#if 0
#define PRB0_TAIL _MMIO(0x2030)
#define PRB0_HEAD _MMIO(0x2034)
@@ -2792,6 +2798,11 @@ static inline bool i915_mmio_reg_valid(i
#define FAULT_VA_HIGH_BITS (0xf << 0)
#define FAULT_GTT_SEL (1 << 4)
+#define GEN12_GFX_TLB_INV_CR _MMIO(0xced8)
+#define GEN12_VD_TLB_INV_CR _MMIO(0xcedc)
+#define GEN12_VE_TLB_INV_CR _MMIO(0xcee0)
+#define GEN12_BLT_TLB_INV_CR _MMIO(0xcee4)
+
#define GEN12_AUX_ERR_DBG _MMIO(0x43f4)
#define FPGA_DBG _MMIO(0x42300)
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -431,6 +431,9 @@ int i915_vma_bind(struct i915_vma *vma,
vma->ops->bind_vma(vma->vm, NULL, vma, cache_level, bind_flags);
}
+ if (vma->obj)
+ set_bit(I915_BO_WAS_BOUND_BIT, &vma->obj->flags);
+
atomic_or(bind_flags, &vma->flags);
return 0;
}
--- a/drivers/gpu/drm/i915/intel_uncore.c
+++ b/drivers/gpu/drm/i915/intel_uncore.c
@@ -724,7 +724,8 @@ void intel_uncore_forcewake_get__locked(
}
static void __intel_uncore_forcewake_put(struct intel_uncore *uncore,
- enum forcewake_domains fw_domains)
+ enum forcewake_domains fw_domains,
+ bool delayed)
{
struct intel_uncore_forcewake_domain *domain;
unsigned int tmp;
@@ -739,7 +740,11 @@ static void __intel_uncore_forcewake_put
continue;
}
- fw_domains_put(uncore, domain->mask);
+ if (delayed &&
+ !(domain->uncore->fw_domains_timer & domain->mask))
+ fw_domain_arm_timer(domain);
+ else
+ fw_domains_put(uncore, domain->mask);
}
}
@@ -760,7 +765,20 @@ void intel_uncore_forcewake_put(struct i
return;
spin_lock_irqsave(&uncore->lock, irqflags);
- __intel_uncore_forcewake_put(uncore, fw_domains);
+ __intel_uncore_forcewake_put(uncore, fw_domains, false);
+ spin_unlock_irqrestore(&uncore->lock, irqflags);
+}
+
+void intel_uncore_forcewake_put_delayed(struct intel_uncore *uncore,
+ enum forcewake_domains fw_domains)
+{
+ unsigned long irqflags;
+
+ if (!uncore->fw_get_funcs)
+ return;
+
+ spin_lock_irqsave(&uncore->lock, irqflags);
+ __intel_uncore_forcewake_put(uncore, fw_domains, true);
spin_unlock_irqrestore(&uncore->lock, irqflags);
}
@@ -802,7 +820,7 @@ void intel_uncore_forcewake_put__locked(
if (!uncore->fw_get_funcs)
return;
- __intel_uncore_forcewake_put(uncore, fw_domains);
+ __intel_uncore_forcewake_put(uncore, fw_domains, false);
}
void assert_forcewakes_inactive(struct intel_uncore *uncore)
--- a/drivers/gpu/drm/i915/intel_uncore.h
+++ b/drivers/gpu/drm/i915/intel_uncore.h
@@ -243,6 +243,8 @@ void intel_uncore_forcewake_get(struct i
enum forcewake_domains domains);
void intel_uncore_forcewake_put(struct intel_uncore *uncore,
enum forcewake_domains domains);
+void intel_uncore_forcewake_put_delayed(struct intel_uncore *uncore,
+ enum forcewake_domains domains);
void intel_uncore_forcewake_flush(struct intel_uncore *uncore,
enum forcewake_domains fw_domains);
Powered by blists - more mailing lists