lib/intel_blt.c: ensure uint64_t result of multiplication
PATCH 1/1 lib/intel_blt.c: ensure uint64_t result of multiplication
Additionally check for overflow.
This should allow to exercise large buffers
ex. xe_exercise_blt -W 16384 -H 16384
--- lib/intel_blt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
lib/intel_blt.c +7 -1
@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping) {struct blt_copy_object *obj; -uint64_t size = width * height * bpp /8 ;uint32_t stride = tiling ==T_LINEAR ? width *4 : width;uint32_t handle; +uint64_t size;igt_assert_f (blt->driver,"Driver isn't set, have you called blt_copy_init()?\n" ); +igt_assert_f ((UINT64_MAX /8 ) >= width && + (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp,"Overflow detected!\n" ); + + size = (uint64_t )width * height * bpp /8 ; + obj =calloc (1 ,sizeof (*obj)); obj->size = size;
Signature · 1 line
2.42.0
Patchwork
CI[igt-dev] ✗ CI.xeBAT: failure for lib/intel_blt.c: ensure uint64_t result of multiplication
== Series Details ==
== Series Details ==
Series: lib/intel_blt.c: ensure uint64_t result of multiplication
URL : https://patchwork.freedesktop.org/series/125241/
State : failure
== Summary ==
CI Bug Log - changes from XEIGT_7543_BAT -> XEIGTPW_10019_BAT
====================================================
Summary
-------
**FAILURE**
Serious unknown changes coming with XEIGTPW_10019_BAT absolutely need to be verified manually.
If you think the reported changes have nothing to do with the changes
introduced in XEIGTPW_10019_BAT, please notify your bug team ([email protected]) to allow them to document this new failure mode, which will reduce false positives in CI.
Participating hosts (4 -> 4)
------------------------------
No changes in participating hosts
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in XEIGTPW_10019_BAT:
### IGT changes ###
#### Possible regressions ####
* igt@kms_frontbuffer_tracking@basic:
- bat-adlp-7: NOTRUN -> [DMESG-FAIL][1]
[1]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
Known issues
------------
Here are the changes found in XEIGTPW_10019_BAT that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@kms_addfb_basic@basic-y-tiled-legacy:
- bat-adlp-7: NOTRUN -> [FAIL][2] ([Intel XE#609]) +2 other tests fail
[2]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
* igt@kms_addfb_basic@invalid-set-prop-any:
- bat-atsm-2: NOTRUN -> [SKIP][3] ([i915#6077]) +33 other tests skip
[3]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_cursor_legacy@basic-flip-before-cursor-legacy:
- bat-atsm-2: NOTRUN -> [SKIP][4] ([Intel XE#782]) +5 other tests skip
[4]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_dsc@dsc-basic:
- bat-atsm-2: NOTRUN -> [SKIP][5] ([Intel XE#784])
[5]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
- bat-adlp-7: NOTRUN -> [SKIP][6] ([Intel XE#423])
[6]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
* igt@kms_flip@basic-flip-vs-modeset:
- bat-atsm-2: NOTRUN -> [SKIP][7] ([Intel XE#541]) +3 other tests skip
[7]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_flip@basic-flip-vs-wf_vblank@a-edp1:
- bat-adlp-7: NOTRUN -> [FAIL][8] ([Intel XE#480]) +1 other test fail
[8]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@kms_flip@[email protected]
* igt@kms_force_connector_basic@force-connector-state:
- bat-atsm-2: NOTRUN -> [SKIP][9] ([Intel XE#540]) +3 other tests skip
[9]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_frontbuffer_tracking@basic:
- bat-atsm-2: NOTRUN -> [SKIP][10] ([Intel XE#783])
[10]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_pipe_crc_basic@compare-crc-sanitycheck-xr24:
- bat-atsm-2: NOTRUN -> [SKIP][11] ([i915#1836]) +6 other tests skip
[11]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_prop_blob@basic:
- bat-atsm-2: NOTRUN -> [SKIP][12] ([Intel XE#780])
[12]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@kms_psr@cursor_plane_move:
- bat-atsm-2: NOTRUN -> [SKIP][13] ([Intel XE#535]) +2 other tests skip
[13]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@kms_psr@cursor_plane_move.html
* igt@xe_compute@compute-square:
- bat-atsm-2: NOTRUN -> [SKIP][14] ([Intel XE#672])
[14]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@xe_evict@evict-beng-small-external:
- bat-adlp-7: NOTRUN -> [SKIP][15] ([Intel XE#261] / [Intel XE#688]) +15 other tests skip
[15]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
* igt@xe_exec_fault_mode@twice-userptr:
- bat-adlp-7: NOTRUN -> [SKIP][16] ([Intel XE#288]) +17 other tests skip
[16]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
* igt@xe_exec_fault_mode@twice-userptr-invalidate-imm:
- bat-atsm-2: NOTRUN -> [SKIP][17] ([Intel XE#288]) +17 other tests skip
[17]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
* igt@xe_huc_copy@huc_copy:
- bat-atsm-2: NOTRUN -> [SKIP][18] ([Intel XE#255])
[18]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@xe_huc_copy@huc_copy.html
* igt@xe_live_ktest@migrate:
- bat-adlp-7: NOTRUN -> [INCOMPLETE][19] ([Intel XE#753])
[19]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
* igt@xe_mmap@vram:
- bat-adlp-7: NOTRUN -> [SKIP][20] ([Intel XE#263])
[20]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-adlp-7/igt@[email protected]
#### Possible fixes ####
* igt@xe_module_load@load:
- bat-atsm-2: [INCOMPLETE][21] -> [PASS][22]
[21]: https://intel-gfx-ci.01.org/tree/intel-xe/IGT_7543/bat-atsm-2/igt@[email protected]
[22]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/bat-atsm-2/igt@[email protected]
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[Intel XE#255]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/255
[Intel XE#261]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/261
[Intel XE#263]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/263
[Intel XE#288]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/288
[Intel XE#423]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/423
[Intel XE#480]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/480
[Intel XE#524]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/524
[Intel XE#535]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/535
[Intel XE#540]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/540
[Intel XE#541]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/541
[Intel XE#609]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/609
[Intel XE#672]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/672
[Intel XE#688]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/688
[Intel XE#753]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/753
[Intel XE#780]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/780
[Intel XE#782]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/782
[Intel XE#783]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/783
[Intel XE#784]: https://gitlab.freedesktop.org/drm/xe/kernel/issues/784
[i915#1836]: https://gitlab.freedesktop.org/drm/intel/issues/1836
[i915#6077]: https://gitlab.freedesktop.org/drm/intel/issues/6077
Build changes
-------------
* IGT: IGT_7543 -> IGTPW_10019
IGTPW_10019: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/index.html
IGT_7543: 688f12831f876590e084e9b13b4d5ab85fe13d51 @ https://gitlab.freedesktop.org/drm/igt-gpu-tools.git
xe-439-cbcd3e95647ff14d508d1a42dce785693446c7f2: cbcd3e95647ff14d508d1a42dce785693446c7f2
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10019/index.html
Patchwork
CI[igt-dev] ✗ Fi.CI.BAT: failure for lib/intel_blt.c: ensure uint64_t result of multiplication
In reply to Marcin Bernatowicz== Series Details ==
== Series Details ==
Series: lib/intel_blt.c: ensure uint64_t result of multiplication
URL : https://patchwork.freedesktop.org/series/125241/
State : failure
== Summary ==
CI Bug Log - changes from IGT_7543 -> IGTPW_10019
====================================================
Summary
-------
**FAILURE**
Serious unknown changes coming with IGTPW_10019 absolutely need to be verified manually.
If you think the reported changes have nothing to do with the changes
introduced in IGTPW_10019, please notify your bug team ([email protected]) to allow them to document this new failure mode, which will reduce false positives in CI.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/index.html
Participating hosts (31 -> 32)
------------------------------
Additional (4): bat-kbl-2 fi-ivb-3770 fi-ilk-650 fi-bsw-n3050
Missing (3): bat-dg2-8 fi-cfl-8109u fi-snb-2520m
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in IGTPW_10019:
### IGT changes ###
#### Possible regressions ####
* igt@gem_exec_parallel@engines@fds:
- fi-skl-guc: [PASS][1] -> [INCOMPLETE][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/IGT_7543/fi-skl-guc/igt@gem_exec_parallel@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/fi-skl-guc/igt@gem_exec_parallel@[email protected]
* igt@i915_module_load@load:
- bat-jsl-3: [PASS][3] -> [INCOMPLETE][4]
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/IGT_7543/bat-jsl-3/igt@[email protected]
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-jsl-3/igt@[email protected]
Known issues
------------
Here are the changes found in IGTPW_10019 that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@fbdev@info:
- bat-kbl-2: NOTRUN -> [SKIP][5] ([fdo#109271] / [i915#1849])
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-kbl-2/igt@[email protected]
* igt@gem_exec_parallel@engines@contexts:
- bat-dg2-9: [PASS][6] -> [INCOMPLETE][7] ([i915#9541])
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/IGT_7543/bat-dg2-9/igt@gem_exec_parallel@[email protected]
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-dg2-9/igt@gem_exec_parallel@[email protected]
* igt@gem_lmem_swapping@parallel-random-engines:
- bat-kbl-2: NOTRUN -> [SKIP][8] ([fdo#109271]) +39 other tests skip
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-kbl-2/igt@[email protected]
* igt@gem_lmem_swapping@random-engines:
- fi-bsw-n3050: NOTRUN -> [SKIP][9] ([fdo#109271]) +18 other tests skip
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/fi-bsw-n3050/igt@[email protected]
* igt@i915_suspend@basic-s2idle-without-i915:
- fi-ivb-3770: NOTRUN -> [DMESG-WARN][10] ([i915#8841]) +6 other tests dmesg-warn
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/fi-ivb-3770/igt@[email protected]
* igt@i915_suspend@basic-s3-without-i915:
- bat-mtlp-8: NOTRUN -> [SKIP][11] ([i915#6645])
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-mtlp-8/igt@[email protected]
* igt@kms_hdmi_inject@inject-audio:
- fi-bsw-n3050: NOTRUN -> [FAIL][12] ([IGT#3])
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/fi-bsw-n3050/igt@[email protected]
* igt@kms_pipe_crc_basic@read-crc-frame-sequence@pipe-d-edp-1:
- bat-rplp-1: [PASS][13] -> [ABORT][14] ([i915#8668])
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/IGT_7543/bat-rplp-1/igt@kms_pipe_crc_basic@[email protected]
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-rplp-1/igt@kms_pipe_crc_basic@[email protected]
* igt@kms_psr@cursor_plane_move:
- fi-ivb-3770: NOTRUN -> [SKIP][15] ([fdo#109271]) +19 other tests skip
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/fi-ivb-3770/igt@kms_psr@cursor_plane_move.html
- fi-ilk-650: NOTRUN -> [SKIP][16] ([fdo#109271]) +19 other tests skip
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/fi-ilk-650/igt@kms_psr@cursor_plane_move.html
#### Possible fixes ####
* igt@i915_selftest@live@requests:
- bat-mtlp-8: [ABORT][17] ([i915#9414]) -> [PASS][18]
[17]: https://intel-gfx-ci.01.org/tree/drm-tip/IGT_7543/bat-mtlp-8/igt@i915_selftest@[email protected]
[18]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-mtlp-8/igt@i915_selftest@[email protected]
* igt@i915_suspend@basic-s2idle-without-i915:
- bat-dg1-5: [INCOMPLETE][19] -> [PASS][20]
[19]: https://intel-gfx-ci.01.org/tree/drm-tip/IGT_7543/bat-dg1-5/igt@[email protected]
[20]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/bat-dg1-5/igt@[email protected]
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[IGT#3]: https://gitlab.freedesktop.org/drm/igt-gpu-tools/issues/3
[fdo#109271]: https://bugs.freedesktop.org/show_bug.cgi?id=109271
[i915#1849]: https://gitlab.freedesktop.org/drm/intel/issues/1849
[i915#6645]: https://gitlab.freedesktop.org/drm/intel/issues/6645
[i915#8668]: https://gitlab.freedesktop.org/drm/intel/issues/8668
[i915#8841]: https://gitlab.freedesktop.org/drm/intel/issues/8841
[i915#9414]: https://gitlab.freedesktop.org/drm/intel/issues/9414
[i915#9541]: https://gitlab.freedesktop.org/drm/intel/issues/9541
Build changes
-------------
* CI: CI-20190529 -> None
* IGT: IGT_7543 -> IGTPW_10019
CI-20190529: 20190529
CI_DRM_13765: 201c8a7bd1f3f415920a2df4b8a8817e973f42fe @ git://anongit.freedesktop.org/gfx-ci/linux
IGTPW_10019: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/index.html
IGT_7543: 688f12831f876590e084e9b13b4d5ab85fe13d51 @ https://gitlab.freedesktop.org/drm/igt-gpu-tools.git
Testlist changes
----------------
-igt@xe_copy_basic@mem-copy-linear-0x3fff
-igt@xe_copy_basic@mem-copy-linear-0x369
-igt@xe_copy_basic@mem-copy-linear-0xfd
-igt@xe_copy_basic@mem-copy-linear-0xfffe
-igt@xe_copy_basic@mem-set-linear-0x3fff
-igt@xe_copy_basic@mem-set-linear-0x369
-igt@xe_copy_basic@mem-set-linear-0xfd
-igt@xe_copy_basic@mem-set-linear-0xfffe
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_10019/index.html
Quoted 4 lines · Marcin Bernatowicz
Additionally check for overflow.
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384
I think it would be good to add a dedicated (sub)test case to test large buffers.
Quoted 14 lines
Signed-off-by Marcin Bernatowicz <[email protected]>--- lib/intel_blt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +0
-0@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping) {struct blt_copy_object *obj;
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1618
Quoted 10 lines
lib/intel_blt.c +4
-1-uint64_t size = width * height * bpp /8 ;uint32_t stride = tiling ==T_LINEAR ? width *4 : width;uint32_t handle; +uint64_t size;igt_assert_f (blt->driver,"Driver isn't set, have you called blt_copy_init()?\n" ); +igt_assert_f ((UINT64_MAX /8 ) >= width && + (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp,"Overflow detected!\n" );
OK, it took a bit for me to parse it... So, we first check if we have enough space to at least fit width, then compare it against height, and then compare all of this to what's left for bpp. Is that correct?
But still, width and height is limited by surface_with and surface_height field that are u14 iirc. I don't think that overflow is possible here?
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1620
Quoted 2 lines
lib/intel_blt.c +2
-0+ + size = (uint64_t )width * height * bpp /8 ;
Cast takes precedence over *, but this line doesn't trigger any warnings, so I think we don't have to add extra ().
All the best,
Karolina
Quoted 4 lines
lib/intel_blt.c +1
-0+ obj =calloc (1 ,sizeof (*obj)); obj->size = size;
Hi,
Quoted 8 lines · Karolina Stolarek
Quoted 4 lines · Marcin Bernatowicz
Additionally check for overflow.
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384I think it would be good to add a dedicated (sub)test case to test large buffers.
Maybe, but it's not the purpose of this patch.
Quoted 18 lines
Quoted 18 lines
Signed-off-by Marcin Bernatowicz <[email protected]>---
lib/intel_blt.c| 8 +++++++-1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +0
-1@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping) { struct blt_copy_object *obj; - uint64_t size = width * height * bpp / 8; uint32_t stride = tiling == T_LINEAR ? width * 4 : width; uint32_t handle;
Quoted 8 lines
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region,
Quoted 6 lines
lib/intel_blt.c +4
-0+ uint64_t size; igt_assert_f(blt->driver, "Driver isn't set, have you called blt_copy_init()?\n"); + igt_assert_f((UINT64_MAX / 8) >= width && + (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp, "Overflow detected!\n");OK, it took a bit for me to parse it... So, we first check if we have enough space to at least fit width, then compare it against height, and then compare all of this to what's left for bpp. Is that correct?
This one is to check if (width * height * bpp / 8) fits in UINT64_MAX and is just a paranoid check.
Quoted 2 lines
But still, width and height is limited by
surface_withandsurface_heightfield that are u14 iirc. I don't think that overflow is possible here?
Where do we limit width and height to u14 ? I see uint32_t is used for width, height and bpp input params.
The current code overflows uint32_t when width = height = 16384 and bpp = 32 => 16384 * 16384 * 32 / 8 => 0 => size == 0 :(
We need (uint64_t) cast to ensure the entire multiplication is done in 64-bit arithmetic.
--
marcin
Quoted 12 lines
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1620
Quoted 3 lines
lib/intel_blt.c +2
-0+ + size = (uint64_t)width * height * bpp / 8;Cast takes precedence over *, but this line doesn't trigger any warnings, so I think we don't have to add extra ().
All the best,
KarolinaQuoted 3 lines
lib/intel_blt.c +1
-0+ obj = calloc(1, sizeof(*obj)); obj->size = size;
Quoted 4 lines · Marcin
Hi,
Quoted 1 line · Karolina Stolarek
On 17.10.2023 16:36, Marcin Bernatowicz wrote:
Quoted 8 lines
Quoted 7 lines
Quoted 4 lines
Additionally check for overflow.
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384I think it would be good to add a dedicated (sub)test case to test large buffers.
Maybe, but it's not the purpose of this patch.
Yes, I agree, but your change suggests that we've missed a test case that should be there. To be clear, this suggestion won't block the patch.
Quoted 28 lines
Quoted 19 lines
Quoted 19 lines
Signed-off-by Marcin Bernatowicz <[email protected]>---
lib/intel_blt.c| 8 +++++++-1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +0
-1@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping) { struct blt_copy_object *obj; - uint64_t size = width * height * bpp / 8; uint32_t stride = tiling == T_LINEAR ? width * 4 : width; uint32_t handle;Quoted 8 lines
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region,
Quoted 6 lines
lib/intel_blt.c +4
-0+ uint64_t size; igt_assert_f(blt->driver, "Driver isn't set, have you called blt_copy_init()?\n"); + igt_assert_f((UINT64_MAX / 8) >= width && + (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp, "Overflow detected!\n");OK, it took a bit for me to parse it... So, we first check if we have enough space to at least fit width, then compare it against height, and then compare all of this to what's left for bpp. Is that correct?
Quoted 2 lines
This one is to check if (width * height * bpp / 8) fits in
UINT64_MAXand is just a paranoid check.
I understand the purpose, I was just checking if it does it right, one step at a time...
Quoted 5 lines
Quoted 3 lines
But still, width and height is limited by
surface_withandsurface_heightfield that are u14 iirc. I don't think that overflow is possible here?Where do we limit width and height to u14 ? I see
uint32_tis used for width, height and bpp input params.
True that we don't limit the output.
Looking at this now, I think we should reject widths and heights that don't fit into u14, because both block copy won't be able to handle it. Fast copy expect pitches, and these are u16. When preparing the command submission, we use helper structs which definitions would truncate the width and height, for example:
dw16 of gen12_block_copy_data_ext:
uint32_t dst_surface_height: BITRANGE(0, 13); uint32_t dst_surface_width: BITRANGE(14, 27);
dw07 of gen12_fast_copy_data:
uint32_t src_pitch: BITRANGE(0, 15);
So, I think, that while we should ensure there's no overflow, we should give folks a heads up if they wish to copy bigger buffers (the latter is more to a note to myself, don't worry). There was a similar issue connected to display, and the only sensible solution was to use render copy that has no such limitations.
All the best,
Karolina
Quoted 18 lines
The current code overflows
uint32_twhen width = height = 16384 and bpp = 32 => 16384 * 16384 * 32 / 8 => 0 => size == 0 :(We need (
uint64_t) cast to ensure the entire multiplication is done in 64-bit arithmetic. >Signature · 1 line
marcinQuoted 12 lines
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1620
Quoted 3 lines
lib/intel_blt.c +2
-0+ + size = (uint64_t)width * height * bpp / 8;Cast takes precedence over *, but this line doesn't trigger any warnings, so I think we don't have to add extra ().
All the best,
KarolinaQuoted 3 lines
lib/intel_blt.c +1
-0+ obj = calloc(1, sizeof(*obj)); obj->size = size;
Hi Marcin,
Quoted 1 line · Marcin Bernatowicz
Additionally check for overflow.
- ^^^^^^^^^^^^
This type was from the start uint64, so imho change subject from:
lib/intel_blt.c: ensure uint64_t result of multiplication ------------ ^^
sidenote: remove ".c"
into:lib/intel_blt: check for overflow in multiplication
and adjust description.
Quoted 3 lines
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384
Please explain - this should fit in 32bit? 16K*16K*32 = 0x40000000 Or do you mean much higher values for W and H?
Quoted 12 lines
Signed-off-by Marcin Bernatowicz <[email protected]>--- lib/intel_blt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +0
-0@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping)
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1616
Quoted 10 lines
lib/intel_blt.c +2
-1{struct blt_copy_object *obj; -uint64_t size = width * height * bpp /8 ;uint32_t stride = tiling ==T_LINEAR ? width *4 : width;uint32_t handle; +uint64_t size;igt_assert_f (blt->driver,"Driver isn't set, have you called blt_copy_init()?\n" ); + igt_assert_f((UINT64_MAX / 8) >= width &&
----------------- ^^^^^^^^^^^^^^
This is not needed, it checks for MAX >= w * 8, while you want size > 0, imho add a second assert after calculating size.
Regards,
Kamil
Quoted 11 lines
lib/intel_blt.c +5
-0+ (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp, "Overflow detected!\n"); + + size = (uint64_t )width * height * bpp /8 ; + obj =calloc (1 ,sizeof (*obj)); obj->size = size;Signature · 1 line
2.42.0
Quoted 6 lines · Kamil Konieczny
Hi Marcin,
Quoted 1 line · Marcin Bernatowicz
Additionally check for overflow.
- ^^^^^^^^^^^^
This type was from the start uint64, so imho change subject from:
Quoted 8 lines
lib/intel_blt.c: ensureuint64_tresult of multiplication ------------ ^^
sidenote: remove ".c"into:
lib/intel_blt: check for overflow in multiplicationand adjust description.
ensure 64-bit arithmetic multiplication ?
Quoted 6 lines
Quoted 4 lines
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384Please explain - this should fit in 32bit? 16K*16K*32 =
0x40000000
Given function blt_create_object(..., uint32_t width, uint32_t height, uint32_t bpp,..) I'm expecting correct uint64_t size calculation.
uint64_t size = width * height * bpp / 8;
Quoted 1 line
Or do you mean much higher values for W and H?
whatever user provides ;)
Quoted 20 lines
Quoted 20 lines
Signed-off-by Marcin Bernatowicz <[email protected]>--- lib/intel_blt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +1
-1@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping) {struct blt_copy_object *obj; -uint64_t size = width * height * bpp /8 ;uint32_t stride = tiling ==T_LINEAR ? width *4 : width;uint32_t handle; +uint64_t size;
Quoted 6 lines
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1616
Quoted 3 lines
lib/intel_blt.c +1
-0igt_assert_f (blt->driver,"Driver isn't set, have you called blt_copy_init()?\n" ); + igt_assert_f((UINT64_MAX / 8) >= width &&----------------- ^^^^^^^^^^^^^^
This is not needed, it checks for MAX >= w * 8, while you want size > 0, imho add a second assert after calculating size.
There is no possibility of uint64_t overflow given uint32_t * uint32_t * uint32_t / 8 multiplication ?
Should I remove this paranoid check?
The most important is (uint64_t) cast in this patch to ensure 64-bit arithmetic.
size > 0 may be an additional check, but do it twice ? (second one after broken ALIGN ;)
--
marcin
Quoted 15 lines
Regards,
KamilQuoted 11 lines
lib/intel_blt.c +5
-0+ (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp, "Overflow detected!\n"); + + size = (uint64_t )width * height * bpp /8 ; + obj =calloc (1 ,sizeof (*obj)); obj->size = size;Signature · 1 line
2.42.0
Hi Marcin,
Quoted 13 lines · Kamil Konieczny
Hi Marcin,
Quoted 1 line · Marcin Bernatowicz
Additionally check for overflow.
- ^^^^^^^^^^^^
This type was from the start uint64, so imho change subject from:
lib/intel_blt.c: ensureuint64_tresult of multiplication ------------ ^^
sidenote: remove ".c"into:
lib/intel_blt: check for overflow in multiplication
Quoted 9 lines
and adjust description.
Quoted 3 lines
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384Please explain - this should fit in 32bit? 16K*16K*32 =
0x40000000Or do you mean much higher values for W and H?
You were right here, sorry.
Quoted 21 lines
Quoted 21 lines
Signed-off-by Marcin Bernatowicz <[email protected]>--- lib/intel_blt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +1
-1@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping) {struct blt_copy_object *obj; -uint64_t size = width * height * bpp /8 ;uint32_t stride = tiling ==T_LINEAR ? width *4 : width;uint32_t handle; +uint64_t size;igt_assert_f (blt->driver,"Driver isn't set, have you called blt_copy_init()?\n" );
Quoted 8 lines
lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1616
Quoted 1 line
lib/intel_blt.c +1
-0+ igt_assert_f((UINT64_MAX / 8) >= width &&----------------- ^^^^^^^^^^^^^^
This is not needed, it checks for MAX >= w * 8, while you want size > 0, imho add a second assert after calculating size.Regards,
Kamil
One more thing, before these asserts you should check that both width and height are not zero.
Regards,
Kamil
Quoted 11 lines
Quoted 11 lines
lib/intel_blt.c +5
-0+ (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp, "Overflow detected!\n"); + + size = (uint64_t )width * height * bpp /8 ; + obj =calloc (1 ,sizeof (*obj)); obj->size = size;Signature · 1 line
2.42.0
Quoted 54 lines · Kamil Konieczny
Hi Marcin,
Quoted 13 lines · Kamil Konieczny
Hi Marcin,
Quoted 1 line · Marcin Bernatowicz
Additionally check for overflow.
- ^^^^^^^^^^^^
This type was from the start uint64, so imho change subject from:
lib/intel_blt.c: ensureuint64_tresult of multiplication ------------ ^^
sidenote: remove ".c"into:
lib/intel_blt: check for overflow in multiplicationQuoted 9 lines
and adjust description.
Quoted 3 lines
This should allow to exercise large buffers
ex.xe_exercise_blt-W 16384 -H 16384Please explain - this should fit in 32bit? 16K*16K*32 =
0x40000000Or do you mean much higher values for W and H?You were right here, sorry.
Quoted 26 lines
Quoted 12 lines
Signed-off-by Marcin Bernatowicz <[email protected]>--- lib/intel_blt.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)lib/intel_blt.c +0
-0@@ -1607,12 +1607,18 @@ blt_create_object(const struct blt_copy_data *blt, uint32_t region, bool create_mapping)lib/intel_blt.c blt_create_object(const struct blt_copy_data *blt, uint32_t region, · line 1616
Quoted 10 lines
lib/intel_blt.c +2
-1{struct blt_copy_object *obj; -uint64_t size = width * height * bpp /8 ;uint32_t stride = tiling ==T_LINEAR ? width *4 : width;uint32_t handle; +uint64_t size;igt_assert_f (blt->driver,"Driver isn't set, have you called blt_copy_init()?\n" ); + igt_assert_f((UINT64_MAX / 8) >= width &&----------------- ^^^^^^^^^^^^^^
This is not needed, it checks for MAX >= w * 8, while you want size > 0, imho add a second assert after calculating size.
Quoted 6 lines
Quoted 3 lines
Regards,
KamilOne more thing, before these asserts you should check that both width and height are not zero.
True, more asserts needed :)
Quoted 15 lines
Regards,
KamilQuoted 11 lines
Quoted 11 lines
lib/intel_blt.c +5
-0+ (UINT64_MAX / width) >= height && + (UINT64_MAX / (width * height)) >= bpp, "Overflow detected!\n"); + + size = (uint64_t )width * height * bpp /8 ; + obj =calloc (1 ,sizeof (*obj)); obj->size = size;Signature · 1 line
2.42.0