Move handle_gpc_gpccs_exception hal to hal.gr.intr
Pass g->ecc.gr.gpccs_ecc_corrected_err_count[gpc].counter and
g->ecc.gr.gpccs_ecc_uncorrected_err_count[gpc].counter variable address
as parameter to function to avoid dereferencing g->ecc variable
inside hal function.
Update g->ops.gr.handle_gpc_gpcss_exception call to
g->ops.gr.intr.handle_gpc_gpcss_exception
JIRA NVGPU-3016
Change-Id: I6cab6428eb6785261f34ca21f2ce055a9995b408
Signed-off-by: Vinod G <vinodg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2087197
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new hal to get_tpc_exception to hal.gr.intr
This hal helps to avoid register read from the
common handle_tpc_exception function. Add a new struct to report the
tpc_exception type back to the common code to handle the exception.
JIRA NVGPU-3016
Change-Id: Ib504ade0b06b85cd38ccf166328784bab072573e
Signed-off-by: Vinod G <vinodg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2085387
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
With hw minimal headers, lot of unwanted hw registers are stripped.
SW needed few updates to use minimal headers:
1. Use stride value to get non zero instance offset:
gr_pri_gpc0_tpc1_tpccs_tpc_activity_0_r() =
gr_pri_gpc0_tpc0_tpccs_tpc_activity_0_r() +
nvgpu_get_litter_value(g, GPU_LIT_TPC_IN_GPC_STRIDE);
gr_pri_be1_becs_be_activity0_r() = gr_pri_be0_becs_be_activity0_r() +
nvgpu_get_litter_value(g, GPU_LIT_ROP_STRIDE);
2. Broadcast registers should not be used for reading status and they should be
used only for broadcast register writes. Removed following register reads
from gm20b register dump:
NV_PGRAPH_PRI_GPCS_TPC0_TPCCS_TPC_ACTIVITY0
NV_PGRAPH_PRI_GPCS_TPC1_TPCCS_TPC_ACTIVITY0
JIRA NVGPU-2917
Change-Id: Ie1359699136c16b67121038024c2318ddd06190c
Signed-off-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2087231
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add a new unit common.gr.obj_ctx which allocates and initializes GR
context. This unit also takes care of creating global golden image
used to initialize every context.
Add private header obj_ctx_priv.h that defines struct
nvgpu_gr_obj_ctx_golden_image
Add public header obj_ctx.h that exposes functions supported by new unit
This unit now exposes below API to allocate and initialize context
nvgpu_gr_obj_ctx_alloc()
Remove below functions from gk20a/gr_gk20a.c and move them to new unit
with below renames
gr_gk20a_fecs_ctx_bind_channel() -> nvgpu_gr_obj_ctx_bind_channel()
gr_gk20a_fecs_ctx_image_save() -> nvgpu_gr_obj_ctx_image_save()
gk20a_init_sw_bundle() -> nvgpu_gr_obj_ctx_alloc_sw_bundle()
gr_gk20a_alloc_gr_ctx() -> nvgpu_gr_obj_ctx_gr_ctx_alloc()
gr_gk20a_init_golden_ctx_image() ->
nvgpu_gr_obj_ctx_alloc_golden_ctx_image()
Use new APIs in gk20a_alloc_obj_ctx() to allocate context
For now this unit includes <nvgpu/gr/gr.h> and some h/w headers.
But they will be removed in follow up patches
Jira NVGPU-1887
Change-Id: Ib95ec1c19c5b74810f85c2feed8fdd63889d3d22
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2087662
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
Reviewed-by: svc-mobile-misra <svc-mobile-misra@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: Vinod Gopalakrishnakurup <vinodg@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Move nvgpu_gr_wait_initialized to a gr.init hal function.
Move to hal function to avoid circular dependencies of headers.
Update nvgpu_gr_wait_initialized call to
g->ops.gr.init.wait_initialized
JIRA NVGPU-3016
Change-Id: Ia2e5f78da8528c76a8d08512151483579f250676
Signed-off-by: Vinod G <vinodg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2085740
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Move chip specific mc code from common/mc to hal/mc.
Replace gk20a_readl/writel with nvgpu_readl/writel
Replace 0xFFFFFFFFU with U32_MAX hash define
Change local variable names to fix checkpatch errors/warnings
Change BUG to WARN
Move defines to header files
Create new defines for hard coded delays
JIRA NVGPU-2041
Change-Id: I3594121a81da37ef58c47e87c45e96441e4cf8c7
Signed-off-by: Seema Khowala <seemaj@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2085268
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
The exec_regops() API was using a u64 for the num_ops parameter. The
lower level APIs used by exec_regops() expect u32s for this value.
Update the interface to use u32.
This eliminates MISRA Rule 10.3 violations for assignment of objects of
different essential or narrower types.
JIRA: NVGPU-3023
Change-Id: I5a2a22916f81d8b3d882d224d07eedffcde1e3ee
Signed-off-by: Philip Elcan <pelcan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084207
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add missing registers and fields related to ECC Corrected/Uncorrected
errors for gv11b. In particular, add ECC Control register and missing
fields for GR, LTC, PMU and FB, in gv11b.
Jira NVGPU-2520
Change-Id: Ic5d2201c6a7383da8151c1ba6ed83aa0743319f6
Signed-off-by: Rajesh Devaraj <rdevaraj@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2070666
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new unit common.gr.setup that provides runtime setup interfaces to
other units outside of GR unit or to OS-specific code
Move zcull setup call to this unit.
New unit now exposes nvgpu_gr_setup_bind_ctxsw_zcull() to setup zcull
This API internally calls common.gr.zcull API nvgpu_gr_zcull_ctx_setup()
Add new hal g->ops.gr.setup.bind_ctxsw_zcull() and remove
g->ops.gr.zcull.bind_ctxsw_zcull()
Remove nvgpu_channel_gr_zcull_setup() from channel unit
Also remove ctx/subctx header includes sicne channel code need not
configure zcull
Remove gm20b_gr_bind_ctxsw_zcull() since binding is done from common
code
Jira NVGPU-1886
Change-Id: I6f04d19a8b8c003734702c5f6780a03ffc89b717
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2086602
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
In this CL, following ctxsw related code is moved to hal gr falcon.
1. gr_gk20a_wait_ctxsw_ready -> gm20b_gr_falcon_wait_ctxsw_ready
2. gr_gk20a_submit_fecs_method_op ->
gm20b_gr_falcon_submit_fecs_method_op
3. gr_gk20a_submit_fecs_sideband_method_op->
gm20b_gr_falcon_submit_fecs_sideband_method_op
Above functions are populated with following gr.falcon hals and called
from the current code as required:
int (*wait_ctxsw_ready)(struct gk20a *g);
int (*submit_fecs_method_op)(struct gk20a *g,
struct fecs_method_op_gk20a op, bool sleepduringwait);
int (*submit_fecs_sideband_method_op)(struct gk20a *g,
struct fecs_method_op_gk20a op);
Following static function also moved to gr_gk20a.c to hal gr falcon.
gr_gk20a_ctx_wait_ucode -> gm20b_gr_falcon_ctx_wait_ucode
JIRA NVGPU-1881
Change-Id: Icb4238dcacaf46ecfcada8bc8dcdeb68b6278bab
Signed-off-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2085189
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Vinod Gopalakrishnakurup <vinodg@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Remove locally defined timeout call in gr and use common timeout
call.
Replace nvgpu_gr_get_idle_timeout with nvgpu_get_poll_timeout function
Replace following defines to
NVGPU_GR_IDLE_CHECK_DEFAULT_US ---> POLL_DELAY_MIN_US
NVGPU_GR_IDLE_CHECK_MAX_US ---> POLL_DELAY_MIN_US
JIRA NVGPU-1885
Change-Id: I4514a9763fe0687680d50704bc9f22677a1a3df9
Signed-off-by: Vinod G <vinodg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2085031
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new hal g->ops.gr.falcon.set_current_ctx_invalid() in hal.gr.falcon
unit to invalidate current_ctx by setting invalid flag in register
gr_fecs_current_ctx_r()
Use new hal in gr_gk20a_init_golden_ctx_image() instead of accessing the
register directly
Define the hal for all supported chips
Jira NVGPU-2961
Change-Id: I756dac505c661ea2754abdbf6934927d1b469be3
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2085032
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
There is nothing h/w specific in gr_gv11b_update_ctxsw_preemption_mode
anymore. Delete it and re-use gp10b specific hal for volta/tu104
Update gr_gp10b_update_ctxsw_preemption_mode to call platform specific
hals if defined
Jira NVGPU-1887
Change-Id: Idae9ebf780b1e76abf847d8b39aa40c0e0560084
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084751
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new hal g->ops.gr.init.gfxp_wfi_timeout() in hal.gr.init unit
to commit gfxp timeout
Define gv11b chip specific operation
Use new hal in gr_gv11b_update_ctxsw_preemption_mode() instead of
directly committing using register accessors
Jira NVGPU-2961
Change-Id: I7694e3128920d9a2856faecf2e3d10a11f0f986e
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084750
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new hal g->ops.gr.init.commit_cbes_reserve() in hal.gr.init unit
to commit cbes reserve
Define gp10b and gv11b chip specific operations
Use new hals in gr_gp10b_update_ctxsw_preemption_mode() and
gr_gv11b_update_ctxsw_preemption_mode() instead of directly committing
using register accessors
Jira NVGPU-2961
Change-Id: Iea2032ea61264c286b1fab46435ff5a84c90d3da
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084749
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Add new hal g->ops.gr.init.commit_ctxsw_spill() in hal.gr.init unit
to commit spill ctxsw buffer
Define gp10b and gv11b operations
Use new hals in gr_gp10b_update_ctxsw_preemption_mode() and
gr_gv11b_update_ctxsw_preemption_mode() instead of directly committing
using register accessors
Jira NVGPU-2961
Change-Id: Iced02d304f12bcb4e78ea31a7728baa04081e325
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084748
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Below hals are used to get preemption buffer sizes
g->ops.gr.get_ctx_spill_size()
g->ops.gr.get_ctx_pagepool_size()
g->ops.gr.get_ctx_betacb_size()
g->ops.gr.get_ctx_attrib_cb_size()
Move them to hal.gr.init unit
Copy over corresponding gp10b/gv11b definitions
Remove pagepool and attrib_cb size hals from gv11b since gv11b can
re-use gp10b hals
Add spill size and betacb size hals for gv100 and tu104 too since
register values are different on those chips
Remove g->ops.gr.init_gfxp_rtv_cb() hal and replace it by
g->ops.gr.init.get_gfxp_rtv_cb_size() which returns the size of RTV
cb size
Jira NVGPU-2961
Change-Id: I3f2f973c120dbfd22067366f87d06b5c9162defb
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084747
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Moved enable/disable HALs from fifo to tsg:
- tsg.enable
- tsg.disable
gk20a_tsg_enable and gv11b_tsg_enable are moved to HAL,
since they are chip specific, even though they do not
directly access chip registers.
Removed vgpu_gv11b_tsg_enable as it was identical to
gv11b_tsg_enable.
Changed gv11b_fifo_locked_abort_runlist_active_tsgs and
gv11b_fifo_teardown_ch_tsg to use tsg.enable HAL instead
of calling directly gk20a_disable_tsg HAL implementation.
Jira NVGPU-2979
Change-Id: I721650c64dcf8cd158652e362292af45df43819f
Signed-off-by: Thomas Fleury <tfleury@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2083156
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Move handle_tex_exception hal function to hal.gr.intr
Move hal function for gm20b and gp10b chips.
Removed the null implementation in gv11b hal.
Modify ops->gr.handle_tex_exception call to
ops->gr.intr.handle_tex_exception
JIRA NVGPU-3016
Change-Id: Ifd88ab6f35301525f7a58e8ccf2f4796dda640bf
Signed-off-by: Vinod G <vinodg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2084387
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Description:
nvgpu_clk_pmupstate is the global structure in clk units. It is declared
in clk.h and all clk units will include clk.h header.
nvgpu_clk_pmupstate struct will have structure pointers to all clk units
and will include genereic function pointers which are used by most clk
units. The reason why the function pointers is defined in this sturct,
and not included inside g->ops is because, these are only clk specific
functions and rest of the driver code is not dependent on this.
Each unit will have init function to allocate memory for its structure
and will initialize its local functions.
Changes:
1) Introduced nvgpu_clk_pmupstate in clk.h file. All the changes needed
to call the above struct from individual clk units.
2) Removed cyclic dependency headers in clk units by calling function
through pointers defined in clk.h.
3) Initialization of each unit is done in respective unit instead of
doing it in clk unit. Added *_init_pmupstate and *_free_pmupstate to
individual clk units.
4) Each unit clean up will be done separately while refactoring that
unit.
NVGPU-1963
NVGPU-2965
Change-Id: Iee79d7a812b62407252636057b104f952c94a229
Signed-off-by: rmylavarapu <rmylavarapu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2033537
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
RC_TYPE_PBDMA_FAULT is the only recovery type for all the pbdma intr
functions. Thus, rc_type variable is changed to a boolean type
in all implementations of handle_pbdma_intr* functions.
"handled" variable is unused and removed from all the implementations of
handle_pbdma_intr* functions.
handle_pbdma_intr* HAL ops are renamed to handle_intr*.
Jira NVGPU-2950
Change-Id: I9605d930225a38ed76f25b6a94cb02d855f522dd
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2083748
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Code for secure/non-secure ctxsw booting spread across gr_gk20a.c
and gr_gm20b.c. With this change this code is move to gr falcon unit.
Ctxsw loading is now supported with 2 supported common functions:
1.Non secure boot:
int nvgpu_gr_falcon_load_ctxsw_ucode(struct gk20a *g);
2.Secure boot:
int nvgpu_gr_falcon_load_secure_ctxsw_ucode(struct gk20a *g);
Now gr ops function "int (*load_ctxsw_ucode)(struct gk20a *g);" is moved to
gr falcon ops and in chip hals it is set with secure/non-secure booting.
Non-secure booting: nvgpu_gr_falcon_load_ctxsw_ucode support ctxsw loading
in 2 methods: bit-banging uode or booting with bootloader
A. Common and hal functions for non-secure bit-banging ctxsw loading:
Common: static void nvgpu_gr_falcon_load_dmem(struct gk20a *g) ->
Hals: void (*load_gpccs_dmem)(struct gk20a *g,i
const u32 *ucode_u32_data, u32 size);
void (*load_fecs_dmem)(struct gk20a *g,
const u32 *ucode_u32_data, u32 size);
Common: static void nvgpu_gr_falcon_load_imem(struct gk20a *g) ->
Hals: void (*load_gpccs_imem)(struct gk20a *g,
const u32 *ucode_u32_data, u32 size);
void (*load_fecs_imem)(struct gk20a *g,
const u32 *ucode_u32_data, u32 size);
Other basic HALs:
void (*configure_fmodel)(struct gk20a *g); -> configure fmodel for ctxsw loading
void (*start_ucode)(struct gk20a *g); -> start running ctxcw ucode
B.Common and hal functions for non-secure ctxsw loading with bootloader
First get the ctxsw ucode using: nvgpu_gr_falcon_init_ctxsw_ucode, then
Common: static void nvgpu_gr_falcon_load_with_bootloader(struct gk20a *g)
void nvgpu_gr_falcon_bind_instblk((struct gk20a *g) ->
Hal: void (*bind_instblk)(struct gk20a *g, struct nvgpu_mem *mem, u64 inst_ptr);
Common: nvgpu_gr_falcon_load_ctxsw_ucode_segments ->
nvgpu_gr_falcon_load_ctxsw_ucode_header ->
nvgpu_gr_falcon_load_ctxsw_ucode_boot for both fecs and gpccs ->
Hals: void (*load_ctxsw_ucode_header)(struct gk20a *g, u32 reg_offset,
u32 boot_signature, u32 addr_code32, u32 addr_data32,
u32 code_size, u32 data_size);
void (*load_ctxsw_ucode_boot)(struct gk20a *g, u64 reg_offset, u32 boot_entry,
u32 addr_load32, u32 blocks, u32 dst);
Other basic HAL to get gpccs start offset:
u32 (*get_gpccs_start_reg_offset)(void);
C.Secure booting is support with gpmu and acr and with following additional
common function in gr falcon.
static void nvgpu_gr_falcon_load_gpccs_with_bootloader(struct gk20a *g) ->
nvgpu_gr_falcon_bind_instblk and nvgpu_gr_falcon_load_ctxsw_ucode_segments
Additional basic hals:
void (*start_gpccs)(struct gk20a *g);
void (*start_fecs)(struct gk20a *g);
Following ops from gr is removed, since it is not required to set by chip hals:
void (*falcon_load_ucode)(struct gk20a *g, u64 addr_base,
struct gk20a_ctxsw_ucode_segments *segments, u32 reg_offset);
Now this is handled by static common function:
static int nvgpu_gr_falcon_copy_ctxsw_ucode_segments( struct gk20a *g,
struct nvgpu_mem *dst, struct gk20a_ctxsw_ucode_segments *segments,
u32 *bootimage, u32 *code, u32 *data)
JIRA NVGPU-1881
Change-Id: I895a03faaf1a21286316befde24765c8b55075cf
Signed-off-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2083388
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
