| 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
| ||
|
Message-ID: <p4x7rodp2qpwfb3hljtbi36mxjdrt5jr3xhad4rebsldlfxart@cz23of57gvrd>
Date: Fri, 15 Nov 2024 09:20:37 +0200
From: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
To: Neil Armstrong <neil.armstrong@...aro.org>
Cc: Akhil P Oommen <quic_akhilpo@...cinc.com>,
Viresh Kumar <vireshk@...nel.org>, Nishanth Menon <nm@...com>, Stephen Boyd <sboyd@...nel.org>,
"Rafael J. Wysocki" <rafael@...nel.org>, Rob Clark <robdclark@...il.com>, Sean Paul <sean@...rly.run>,
Konrad Dybcio <konradybcio@...nel.org>, Abhinav Kumar <quic_abhinavk@...cinc.com>,
Marijn Suijten <marijn.suijten@...ainline.org>, David Airlie <airlied@...il.com>,
Simona Vetter <simona@...ll.ch>, Bjorn Andersson <andersson@...nel.org>,
Rob Herring <robh@...nel.org>, Krzysztof Kozlowski <krzk+dt@...nel.org>,
Conor Dooley <conor+dt@...nel.org>, Connor Abbott <cwabbott0@...il.com>, linux-pm@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-arm-msm@...r.kernel.org, dri-devel@...ts.freedesktop.org,
freedreno@...ts.freedesktop.org, devicetree@...r.kernel.org
Subject: Re: [PATCH RFC 3/8] drm/msm: adreno: add plumbing to generate
bandwidth vote table for GMU
On Wed, Nov 13, 2024 at 04:48:29PM +0100, Neil Armstrong wrote:
> The Adreno GMU Management Unit (GMU) can also scale DDR Bandwidth along
> the Frequency and Power Domain level, but by default we leave the
> OPP core scale the interconnect ddr path.
>
> In order to get the vote values to be used by the GPU Management
> Unit (GMU), we need to parse all the possible OPP Bandwidths and
> create a vote value to be send to the appropriate Bus Control
> Modules (BCMs) declared in the GPU info struct.
>
> The vote array will be used to dynamically generate the GMU bw_table
> sent during the GMU power-up.
>
> Signed-off-by: Neil Armstrong <neil.armstrong@...aro.org>
> ---
> drivers/gpu/drm/msm/adreno/a6xx_gmu.c | 163 ++++++++++++++++++++++++++++++++++
> drivers/gpu/drm/msm/adreno/a6xx_gmu.h | 12 +++
> drivers/gpu/drm/msm/adreno/a6xx_gpu.h | 1 +
> 3 files changed, 176 insertions(+)
>
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
> index 14db7376c712d19446b38152e480bd5a1e0a5198..504a7c5d5a9df4c787951f2ae3a69d566d205ad5 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
> @@ -9,6 +9,7 @@
> #include <linux/pm_domain.h>
> #include <linux/pm_opp.h>
> #include <soc/qcom/cmd-db.h>
> +#include <soc/qcom/tcs.h>
> #include <drm/drm_gem.h>
>
> #include "a6xx_gpu.h"
> @@ -1287,6 +1288,119 @@ static int a6xx_gmu_memory_probe(struct a6xx_gmu *gmu)
> return 0;
> }
>
> +struct a6xx_bcm_data {
> + u32 buswidth;
> + unsigned int unit;
> + unsigned int width;
In bits?
> + unsigned int vcd;
What is this?
> + bool fixed;
What does it mean?
> + unsigned int perfmode;
> + unsigned int perfmode_bw;
> +};
> +
> +struct bcm_db {
> + __le32 unit;
> + __le16 width;
> + u8 vcd;
> + u8 reserved;
> +};
> +
> +static int a6xx_gmu_rpmh_get_bcm_data(const struct a6xx_bcm *bcm,
> + struct a6xx_bcm_data *bcm_data)
Is there a reason to copy CMD DB and BCM data to the interim
representation instead of using those directly?
> +{
> + const struct bcm_db *data;
> + size_t count;
> +
> + data = cmd_db_read_aux_data(bcm->name, &count);
> + if (IS_ERR(data))
> + return PTR_ERR(data);
> +
> + if (!count)
> + return -EINVAL;
> +
> + bcm_data->unit = le32_to_cpu(data->unit);
> + bcm_data->width = le16_to_cpu(data->width);
> + bcm_data->vcd = data->vcd;
> + bcm_data->fixed = bcm->fixed;
> + bcm_data->perfmode = bcm->perfmode;
> + bcm_data->perfmode_bw = bcm->perfmode_bw;
> + bcm_data->buswidth = bcm->buswidth;
> +
> + return 0;
> +}
> +
> +static void a6xx_gmu_rpmh_calc_bw_vote(struct a6xx_bcm_data *bcms,
> + int count, u32 bw, u32 *data)
> +{
> + int i;
> +
> + for (i = 0; i < count; i++) {
> + bool valid = true;
> + bool commit = false;
> + u64 peak, y;
> +
> + if (i == count - 1 || bcms[i].vcd != bcms[i + 1].vcd)
> + commit = true;
> +
> + if (bcms[i].fixed) {
> + if (!bw)
> + data[i] = BCM_TCS_CMD(commit, false, 0x0, 0x0);
> + else
> + data[i] = BCM_TCS_CMD(commit, true, 0x0,
> + bw >= bcms[i].perfmode_bw ?
> + bcms[i].perfmode : 0x0);
> + continue;
> + }
> +
> + /* Multiple the bandwidth by the width of the connection */
... and divide by the bus width. However it's not clear why you are
multiplying bandwidth (bits or bytes per second) with the width
(probably also bits?). Or is it not a width but the number of paths
between units?
> + peak = (u64)bw * bcms[i].width;
> + do_div(peak, bcms[i].buswidth);
> +
> + /* Input bandwidth value is in KBps */
Input or OPP / Interconnect?
> + y = peak * 1000ULL;
> + do_div(y, bcms[i].unit);
> +
> + /*
> + * If a bandwidth value was specified but the calculation ends
> + * rounding down to zero, set a minimum level
> + */
> + if (bw && y == 0)
> + y = 1;
Is it a real usecase or just a safety net? If the bandwidth ends up
being very low, maybe we should warn the users about it?
> +
> + y = min_t(u64, y, BCM_TCS_CMD_VOTE_MASK);
> + if (!y)
> + valid = false;
This can probably be coupled with the previous condition.
> +
> + data[i] = BCM_TCS_CMD(commit, valid, y, y);
> + }
> +}
> +
> +static int a6xx_gmu_rpmh_bw_votes_init(const struct a6xx_info *info, struct a6xx_gmu *gmu)
> +{
> + struct a6xx_bcm_data bcms[3];
> + unsigned int bcm_count = 0;
> + int ret, index;
> +
> + /* Retrieve BCM data from cmd-db and merge with a6xx_info bcm table */
> + for (index = 0; index < 3; index++) {
Magic number 3.
> + if (!info->bcm[index].name)
> + continue;
> +
> + ret = a6xx_gmu_rpmh_get_bcm_data(&info->bcm[index], &bcms[index]);
> + if (ret)
> + return ret;
> +
> + ++bcm_count;
> + }
> +
> + /* Generate BCM votes values for each bandwidth & bcm */
> + for (index = 0; index < gmu->nr_gpu_bws; index++)
> + a6xx_gmu_rpmh_calc_bw_vote(bcms, bcm_count, gmu->gpu_bw_table[index],
> + gmu->gpu_bw_votes[index]);
> +
> + return 0;
> +}
> +
> /* Return the 'arc-level' for the given frequency */
> static unsigned int a6xx_gmu_get_arc_level(struct device *dev,
> unsigned long freq)
> @@ -1390,12 +1504,15 @@ static int a6xx_gmu_rpmh_arc_votes_init(struct device *dev, u32 *votes,
> * The GMU votes with the RPMh for itself and on behalf of the GPU but we need
> * to construct the list of votes on the CPU and send it over. Query the RPMh
> * voltage levels and build the votes
> + * The GMU can also vote for DDR interconnects, use the OPP bandwidth entries
> + * and BCM parameters to build the votes.
> */
>
> static int a6xx_gmu_rpmh_votes_init(struct a6xx_gmu *gmu)
> {
> struct a6xx_gpu *a6xx_gpu = container_of(gmu, struct a6xx_gpu, gmu);
> struct adreno_gpu *adreno_gpu = &a6xx_gpu->base;
> + const struct a6xx_info *info = adreno_gpu->info->a6xx;
> struct msm_gpu *gpu = &adreno_gpu->base;
> int ret;
>
> @@ -1407,6 +1524,10 @@ static int a6xx_gmu_rpmh_votes_init(struct a6xx_gmu *gmu)
> ret |= a6xx_gmu_rpmh_arc_votes_init(gmu->dev, gmu->cx_arc_votes,
> gmu->gmu_freqs, gmu->nr_gmu_freqs, "cx.lvl");
>
> + /* Build the interconnect votes */
> + if (adreno_gpu->info->quirks & ADRENO_QUIRK_GMU_BW_VOTE)
> + ret |= a6xx_gmu_rpmh_bw_votes_init(info, gmu);
> +
> return ret;
> }
>
> @@ -1442,6 +1563,38 @@ static int a6xx_gmu_build_freq_table(struct device *dev, unsigned long *freqs,
> return index;
> }
>
> +static int a6xx_gmu_build_bw_table(struct device *dev, unsigned long *bandwidths,
> + u32 size)
> +{
> + int count = dev_pm_opp_get_opp_count(dev);
> + struct dev_pm_opp *opp;
> + int i, index = 0;
> + unsigned int bandwidth = 1;
> +
> + /*
> + * The OPP table doesn't contain the "off" bandwidth level so we need to
> + * add 1 to the table size to account for it
> + */
> +
> + if (WARN(count + 1 > size,
> + "The GMU bandwidth table is being truncated\n"))
> + count = size - 1;
> +
> + /* Set the "off" bandwidth */
> + bandwidths[index++] = 0;
> +
> + for (i = 0; i < count; i++) {
> + opp = dev_pm_opp_find_bw_ceil(dev, &bandwidth, 0);
> + if (IS_ERR(opp))
> + break;
> +
> + dev_pm_opp_put(opp);
> + bandwidths[index++] = bandwidth++;
> + }
> +
> + return index;
> +}
> +
> static int a6xx_gmu_pwrlevels_probe(struct a6xx_gmu *gmu)
> {
> struct a6xx_gpu *a6xx_gpu = container_of(gmu, struct a6xx_gpu, gmu);
> @@ -1472,6 +1625,16 @@ static int a6xx_gmu_pwrlevels_probe(struct a6xx_gmu *gmu)
>
> gmu->current_perf_index = gmu->nr_gpu_freqs - 1;
>
> + /*
> + * The GMU also handles GPU Interconnect Votes so build a list
> + * of DDR bandwidths from the GPU OPP table
> + */
> + if (adreno_gpu->info->quirks & ADRENO_QUIRK_GMU_BW_VOTE)
> + gmu->nr_gpu_bws = a6xx_gmu_build_bw_table(&gpu->pdev->dev,
> + gmu->gpu_bw_table, ARRAY_SIZE(gmu->gpu_bw_table));
> +
> + gmu->current_perf_index = gmu->nr_gpu_freqs - 1;
> +
> /* Build the list of RPMh votes that we'll send to the GMU */
> return a6xx_gmu_rpmh_votes_init(gmu);
> }
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gmu.h b/drivers/gpu/drm/msm/adreno/a6xx_gmu.h
> index b4a79f88ccf45cfe651c86d2a9da39541c5772b3..95c632d8987a517f067c48c61c6c06b9a4f61fc0 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gmu.h
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gmu.h
> @@ -19,6 +19,14 @@ struct a6xx_gmu_bo {
> u64 iova;
> };
>
> +struct a6xx_bcm {
> + char *name;
> + unsigned int buswidth;
> + bool fixed;
> + unsigned int perfmode;
> + unsigned int perfmode_bw;
> +};
> +
> /*
> * These define the different GMU wake up options - these define how both the
> * CPU and the GMU bring up the hardware
> @@ -82,6 +90,10 @@ struct a6xx_gmu {
> unsigned long gpu_freqs[16];
> u32 gx_arc_votes[16];
>
> + int nr_gpu_bws;
> + unsigned long gpu_bw_table[16];
> + u32 gpu_bw_votes[16][3];
Is it is the same magic 16 as we have few lines above or is this 16 a
different magic 16? And also 3 is a pure dark secret.
> +
> int nr_gmu_freqs;
> unsigned long gmu_freqs[4];
> u32 cx_arc_votes[4];
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.h b/drivers/gpu/drm/msm/adreno/a6xx_gpu.h
> index 4aceffb6aae89c781facc2a6e4a82b20b341b6cb..d779d700120cbd974ee87a67214739b1d85156e2 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.h
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.h
> @@ -44,6 +44,7 @@ struct a6xx_info {
> u32 gmu_chipid;
> u32 gmu_cgc_mode;
> u32 prim_fifo_threshold;
> + const struct a6xx_bcm bcm[3];
> };
>
> struct a6xx_gpu {
>
> --
> 2.34.1
>
--
With best wishes
Dmitry
Powered by blists - more mailing lists