Changes for introducing the quick test on the device memory range and also a test of detailed validation for each addr of the range with read and write.
Detailed testing is optionally enabled with a modparam i915.memtest=1
And third patch fixes the driver accessible stolen memory.
v2: Adding a wrapper for the memtest [Chris]
Chris Wilson (3): drm/i915: Exclude reserved stolen from driver use drm/i915: Sanitycheck device iomem on probe drm/i915: Test all device memory on probing
drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 3 + drivers/gpu/drm/i915/i915_params.c | 3 + drivers/gpu/drm/i915/i915_params.h | 1 + drivers/gpu/drm/i915/intel_memory_region.c | 128 +++++++++++++++++++++ 4 files changed, 135 insertions(+)
From: Chris Wilson chris@chris-wilson.co.uk
Remove the portion of stolen memory reserved for private use from driver access.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk cc: Matthew Auld matthew.auld@intel.com Signed-off-by: Ramalingam C ramalingam.c@intel.com Reviewed-by: Matthew Auld matthew.auld@intel.com --- drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 3 +++ 1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c index bce03d74a0b4..6ea3ca21cdf3 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c @@ -488,6 +488,9 @@ static int i915_gem_init_stolen(struct intel_memory_region *mem) return 0; }
+ /* Exclude the reserved region from driver use */ + mem->region.end = reserved_base - 1; + /* It is possible for the reserved area to end before the end of stolen * memory, so just consider the start. */ reserved_total = stolen_top - reserved_base;
On Wed, Dec 08, 2021 at 08:27:58PM +0530, Ramalingam C wrote:
From: Chris Wilson chris@chris-wilson.co.uk
Remove the portion of stolen memory reserved for private use from driver access.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk cc: Matthew Auld matthew.auld@intel.com Signed-off-by: Ramalingam C ramalingam.c@intel.com Reviewed-by: Matthew Auld matthew.auld@intel.com
drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 3 +++ 1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c index bce03d74a0b4..6ea3ca21cdf3 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c @@ -488,6 +488,9 @@ static int i915_gem_init_stolen(struct intel_memory_region *mem) return 0; }
- /* Exclude the reserved region from driver use */
- mem->region.end = reserved_base - 1;
- /* It is possible for the reserved area to end before the end of stolen
reserved_total = stolen_top - reserved_base;
- memory, so just consider the start. */
-- 2.20.1
Hi Ram,
On Wed, Dec 08, 2021 at 08:27:58PM +0530, Ramalingam C wrote:
From: Chris Wilson chris@chris-wilson.co.uk
Remove the portion of stolen memory reserved for private use from driver access.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk cc: Matthew Auld matthew.auld@intel.com Signed-off-by: Ramalingam C ramalingam.c@intel.com Reviewed-by: Matthew Auld matthew.auld@intel.com
Reviewed-by: Andi Shyti andi.shyti@linux.intel.com
Thanks, Andi
From: Chris Wilson chris@chris-wilson.co.uk
As we setup the memory regions for the device, give each a quick test to verify that we can read and write to the full iomem range. This ensures that our physical addressing for the device's memory is correct, and some reassurance that the memory is functional.
v2: wrapper for memtest [Chris]
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk Cc: Matthew Auld matthew.auld@intel.com Signed-off-by: Ramalingam C ramalingam.c@intel.com Reviewed-by: Matthew Auld matthew.auld@intel.com --- drivers/gpu/drm/i915/intel_memory_region.c | 117 +++++++++++++++++++++ 1 file changed, 117 insertions(+)
diff --git a/drivers/gpu/drm/i915/intel_memory_region.c b/drivers/gpu/drm/i915/intel_memory_region.c index b43121609e25..ed65b67e2d14 100644 --- a/drivers/gpu/drm/i915/intel_memory_region.c +++ b/drivers/gpu/drm/i915/intel_memory_region.c @@ -3,6 +3,8 @@ * Copyright © 2019 Intel Corporation */
+#include <linux/prandom.h> + #include "intel_memory_region.h" #include "i915_drv.h" #include "i915_ttm_buddy_manager.h" @@ -29,6 +31,99 @@ static const struct { }, };
+static int __iopagetest(struct intel_memory_region *mem, + u8 __iomem *va, int pagesize, + u8 value, resource_size_t offset, + const void *caller) +{ + int byte = prandom_u32_max(pagesize); + u8 result[3]; + + memset_io(va, value, pagesize); /* or GPF! */ + wmb(); + + result[0] = ioread8(va); + result[1] = ioread8(va + byte); + result[2] = ioread8(va + pagesize - 1); + if (memchr_inv(result, value, sizeof(result))) { + dev_err(mem->i915->drm.dev, + "Failed to read back from memory region:%pR at [%pa + %pa] for %ps; wrote %x, read (%x, %x, %x)\n", + &mem->region, &mem->io_start, &offset, caller, + value, result[0], result[1], result[2]); + return -EINVAL; + } + + return 0; +} + +static int iopagetest(struct intel_memory_region *mem, + resource_size_t offset, + const void *caller) +{ + const u8 val[] = { 0x0, 0xa5, 0xc3, 0xf0 }; + void __iomem *va; + int err; + int i; + + va = ioremap_wc(mem->io_start + offset, PAGE_SIZE); + if (!va) { + dev_err(mem->i915->drm.dev, + "Failed to ioremap memory region [%pa + %px] for %ps\n", + &mem->io_start, &offset, caller); + return -EFAULT; + } + + for (i = 0; i < ARRAY_SIZE(val); i++) { + err = __iopagetest(mem, va, PAGE_SIZE, val[i], offset, caller); + if (err) + break; + + err = __iopagetest(mem, va, PAGE_SIZE, ~val[i], offset, caller); + if (err) + break; + } + + iounmap(va); + return err; +} + +static resource_size_t random_page(resource_size_t last) +{ + /* Limited to low 44b (16TiB), but should suffice for a spot check */ + return prandom_u32_max(last >> PAGE_SHIFT) << PAGE_SHIFT; +} + +static int iomemtest(struct intel_memory_region *mem, const void *caller) +{ + resource_size_t last = resource_size(&mem->region) - PAGE_SIZE; + int err; + + /* + * Quick test to check read/write access to the iomap (backing store). + * + * Write a byte, read it back. If the iomapping fails, we expect + * a GPF preventing further execution. If the backing store does not + * exist, the read back will return garbage. We check a couple of pages, + * the first and last of the specified region to confirm the backing + * store + iomap does cover the entire memory region; and we check + * a random offset within as a quick spot check for bad memory. + */ + + err = iopagetest(mem, 0, caller); + if (err) + return err; + + err = iopagetest(mem, last, caller); + if (err) + return err; + + err = iopagetest(mem, random_page(last), caller); + if (err) + return err; + + return 0; +} + struct intel_memory_region * intel_memory_region_lookup(struct drm_i915_private *i915, u16 class, u16 instance) @@ -90,6 +185,21 @@ void intel_memory_region_debug(struct intel_memory_region *mr, &mr->total, &mr->avail); }
+static int intel_memory_region_memtest(struct intel_memory_region *mem, + void *caller) +{ + struct drm_i915_private *i915 = mem->i915; + int err = 0; + + if (!mem->io_start) + return 0; + + if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) + err = iomemtest(mem, caller); + + return err; +} + struct intel_memory_region * intel_memory_region_create(struct drm_i915_private *i915, resource_size_t start, @@ -126,8 +236,15 @@ intel_memory_region_create(struct drm_i915_private *i915, goto err_free; }
+ err = intel_memory_region_memtest(mem, (void *)_RET_IP_); + if (err) + goto err_release; + return mem;
+err_release: + if (mem->ops->release) + mem->ops->release(mem); err_free: kfree(mem); return ERR_PTR(err);
Hi Ram,
+static int intel_memory_region_memtest(struct intel_memory_region *mem,
void *caller)+{
- struct drm_i915_private *i915 = mem->i915;
- int err = 0;
- if (!mem->io_start)
return 0;- if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM))
err = iomemtest(mem, caller);
I don't understand the debugging part of the iomemtest, if memory is failing we wouldn't relise if DEBUG_GEM is not enabled.
In any case,
Reviewed-by: Andi Shyti andi.shyti@linux.intel.com
Andi
On 2021-12-08 at 17:23:26 +0200, Andi Shyti wrote:
Hi Ram,
+static int intel_memory_region_memtest(struct intel_memory_region *mem,
void *caller)+{
- struct drm_i915_private *i915 = mem->i915;
- int err = 0;
- if (!mem->io_start)
return 0;- if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM))
err = iomemtest(mem, caller);I don't understand the debugging part of the iomemtest, if memory is failing we wouldn't relise if DEBUG_GEM is not enabled.
Thanks for the review Andi!
Since this is in the system init path, running the time taking mem test only (on demand basis) when the DEBUG_GEM is set or mod_param is set as per the next patch.
Ram
In any case,
Reviewed-by: Andi Shyti andi.shyti@linux.intel.com
Andi
From: Chris Wilson chris@chris-wilson.co.uk
This extends the previous sanitychecking of device memory to read/write all the memory on the device during the device probe, ala memtest86, as an optional module parameter: i915.memtest=1. This is not expected to be fast, but a reasonably thorough verfification that the device memory is accessible and doesn't return bit errors.
v2: Rebased.
Suggested-by: Matthew Auld matthew.auld@intel.com Signed-off-by: Chris Wilson chris@chris-wilson.co.uk Cc: Matthew Auld matthew.auld@intel.com Signed-off-by: Ramalingam C ramalingam.c@intel.com Reviewed-by: Matthew Auld matthew.auld@intel.com --- drivers/gpu/drm/i915/i915_params.c | 3 ++ drivers/gpu/drm/i915/i915_params.h | 1 + drivers/gpu/drm/i915/intel_memory_region.c | 35 ++++++++++++++-------- 3 files changed, 27 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c index e07f4cfea63a..525ae832aa9a 100644 --- a/drivers/gpu/drm/i915/i915_params.c +++ b/drivers/gpu/drm/i915/i915_params.c @@ -140,6 +140,9 @@ i915_param_named_unsafe(invert_brightness, int, 0400, i915_param_named(disable_display, bool, 0400, "Disable display (default: false)");
+i915_param_named(memtest, bool, 0400, + "Perform a read/write test of all device memory on module load (default: off)"); + i915_param_named(mmio_debug, int, 0400, "Enable the MMIO debug code for the first N failures (default: off). " "This may negatively affect performance."); diff --git a/drivers/gpu/drm/i915/i915_params.h b/drivers/gpu/drm/i915/i915_params.h index 8d725b64592d..c9d53ff910a0 100644 --- a/drivers/gpu/drm/i915/i915_params.h +++ b/drivers/gpu/drm/i915/i915_params.h @@ -64,6 +64,7 @@ struct drm_printer; param(char *, guc_firmware_path, NULL, 0400) \ param(char *, huc_firmware_path, NULL, 0400) \ param(char *, dmc_firmware_path, NULL, 0400) \ + param(bool, memtest, false, 0400) \ param(int, mmio_debug, -IS_ENABLED(CONFIG_DRM_I915_DEBUG_MMIO), 0600) \ param(int, edp_vswing, 0, 0400) \ param(unsigned int, reset, 3, 0600) \ diff --git a/drivers/gpu/drm/i915/intel_memory_region.c b/drivers/gpu/drm/i915/intel_memory_region.c index ed65b67e2d14..bab5b5caa6d8 100644 --- a/drivers/gpu/drm/i915/intel_memory_region.c +++ b/drivers/gpu/drm/i915/intel_memory_region.c @@ -93,9 +93,12 @@ static resource_size_t random_page(resource_size_t last) return prandom_u32_max(last >> PAGE_SHIFT) << PAGE_SHIFT; }
-static int iomemtest(struct intel_memory_region *mem, const void *caller) +static int iomemtest(struct intel_memory_region *mem, + bool test_all, + const void *caller) { resource_size_t last = resource_size(&mem->region) - PAGE_SIZE; + resource_size_t page; int err;
/* @@ -109,17 +112,25 @@ static int iomemtest(struct intel_memory_region *mem, const void *caller) * a random offset within as a quick spot check for bad memory. */
- err = iopagetest(mem, 0, caller); - if (err) - return err; + if (test_all) { + for (page = 0; page <= last; page += PAGE_SIZE) { + err = iopagetest(mem, page, caller); + if (err) + return err; + } + } else { + err = iopagetest(mem, 0, caller); + if (err) + return err;
- err = iopagetest(mem, last, caller); - if (err) - return err; + err = iopagetest(mem, last, caller); + if (err) + return err;
- err = iopagetest(mem, random_page(last), caller); - if (err) - return err; + err = iopagetest(mem, random_page(last), caller); + if (err) + return err; + }
return 0; } @@ -194,8 +205,8 @@ static int intel_memory_region_memtest(struct intel_memory_region *mem, if (!mem->io_start) return 0;
- if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) - err = iomemtest(mem, caller); + if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) || i915->params.memtest) + err = iomemtest(mem, i915->params.memtest, caller);
return err; }
Hi Ram and Chris,
param(char *, guc_firmware_path, NULL, 0400) \ param(char *, huc_firmware_path, NULL, 0400) \ param(char *, dmc_firmware_path, NULL, 0400) \
- param(bool, memtest, false, 0400) \
this partially answers my previous question...
[...]
- if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM))
err = iomemtest(mem, caller);
- if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) || i915->params.memtest)
err = iomemtest(mem, i915->params.memtest, caller);
... but still I am missing the debugging part.
Reviewed-by: Andi Shyti andi.shyti@linux.intel.com
Andi
dri-devel@lists.freedesktop.org
-
Andi Shyti -
Ramalingam C