Message ID | cover.1566570260.git.leonard.crestez@nxp.com (mailing list archive) |
---|---|
Headers | show |
Series | interconnect: Add imx support via devfreq | expand |
On 23.08.2019 17:37, Leonard Crestez wrote: > This series add imx support for interconnect via devfreq: the ICC > framework is used to aggregate requests from devices and then those are > converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. > > Since there is no single devicetree node that can represent the "interconnect" > new API is added to allow individual devfreq nodes to act as parsing proxies > all mapping to a single soc-level icc provider. This is still RFC > because of this Any comments? I made a lot of changes relative to previous versions, most of them solely to avoid adding a virtual node in DT bindings. The only current interconnect provider implementation is for qcom and it uses a firmware node as the provider node (with #interconnect-cells). However there is no obvious equivalent of that for imx and many other SOCs. On imx there are multiple pieces of scalable fabric which can be defined in DT as devfreq devices and it sort of makes sense to add #interconnect-cells to those. However when it comes to describing the SOC interconnect graph it's much more convenient to have a single per-SOC platform driver. My solution is to add an "icc_proxy" API so that a single platform-level interconnect provider can be referenced through multiple DT nodes. Does this make sense? The implementation is not very pretty, the interconnect platform devices ends up enumerating the entire devicetree in order to find proxies. Right now the interconnect API use a relatively standard split between consumer and "provider" but I think it might make sense to have a separate abstractions for "graph" and "midnode". A "midnode" could act as a DT proxy if there is no single representation of the "interconnect" and it could support custom scaling for itself (with the default being scaling kbps into MIN_FREQ). There are also other options: * Pick one "main" bus and bless it as the "interconnect provider". I want to represent buses as devfreq devices so I would have to call from devfreq to ICC for registration somehow. * Maybe the "no virtual device" rule could be relaxed for the interconnect subsystem? -- Regards, Leonard
On 23.08.19 16:36, Leonard Crestez wrote: > This series add imx support for interconnect via devfreq: the ICC > framework is used to aggregate requests from devices and then those are > converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. > > The devfreq parts are posted separately, this series only intersects in > devicetree: https://patchwork.kernel.org/cover/11104113/ > > Since there is no single devicetree node that can represent the "interconnect" > new API is added to allow individual devfreq nodes to act as parsing proxies > all mapping to a single soc-level icc provider. This is still RFC > because this > > The rest of the changes are small and deal with review comments. > hi Leonard, When I run https://github.com/cdleonard/linux/tree/next_imx_busfreq and https://github.com/cdleonard/arm-trusted-firmware/tree/imx_2.0.y_busfreq on imx8mq, probe() fails: [ 1.082847] imx-ddrc-devfreq 3d400000.dram-controller: failed to init firmware freq info: -19 [ 1.091434] imx-ddrc-devfreq: probe of 3d400000.dram-controller rejects match -19 in imx_ddrc_init_freq_info()'s check: if (priv->freq_count <= 0 || priv->freq_count > IMX_DDRC_MAX_FREQ_COUNT) That would indicate that I'm missing something in ATF? I'm pretty sure I'm running your tree though. Does anything come to your mind what I might be doing wrong? Am I running your "correct" linux tree? Remember I'm on imx8mq, so I don't exactly test this RFC of yours. Also, how are your plans to rebase / update your ATF tree? And since there's a lot in there: what additions are necessary for this devfreq to work? Lastly, how do you test? Is /sys/class/devfreq supposted to get populated? thanks for your work! martin
On 21.09.2019 09:07, Martin Kepplinger wrote: > On 23.08.19 16:36, Leonard Crestez wrote: >> This series add imx support for interconnect via devfreq: the ICC >> framework is used to aggregate requests from devices and then those are >> converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. >> >> The devfreq parts are posted separately, this series only intersects in >> devicetree. >> >> Since there is no single devicetree node that can represent the "interconnect" >> new API is added to allow individual devfreq nodes to act as parsing proxies >> all mapping to a single soc-level icc provider. This is still RFC >> because this >> >> The rest of the changes are small and deal with review comments. > on imx8mq, probe() fails: > > [ 1.082847] imx-ddrc-devfreq 3d400000.dram-controller: failed to init > firmware freq info: -19 > [ 1.091434] imx-ddrc-devfreq: probe of 3d400000.dram-controller > rejects match -19 > > in imx_ddrc_init_freq_info()'s check: > > if (priv->freq_count <= 0 || priv->freq_count > IMX_DDRC_MAX_FREQ_COUNT) > > That would indicate that I'm missing something in ATF? I'm pretty sure > I'm running your tree though. What is your board and uboot version? I tested on imx8mq-evk (SOC Rev B1) with NXP uboot. For example this uboot release works: https://source.codeaurora.org/external/imx/uboot-imx/log/?h=imx_v2019.04_4.19.35_1.0.0 It is uboot which trains DDR for multiple frequencies and then passes that info to ATF. I'm not sure about the steps required to enable this for 3rd-party boards, should be same as for busfreq from NXP tree. Getting this to work on a 3rd-party board would be interesting. > Does anything come to your mind what I might be doing wrong? Am I > running your "correct" linux tree? Remember I'm on imx8mq, so I don't > exactly test this RFC of yours. > > Also, how are your plans to rebase / update your ATF tree? The ATF changes will show up in a future release of NXP ATF branch, when that happens I will drop my branch. NXP ATF releases are on CAF: https://source.codeaurora.org/external/imx/imx-atf/ > And since there's a lot in there: what additions are necessary for this > devfreq to work? devfreq imx support here: https://patchwork.kernel.org/cover/11104113/ Interconnect support also needs PM QoS support for devfreq: https://patchwork.kernel.org/cover/11157649/ > Lastly, how do you test? Is /sys/class/devfreq supposted to get populated? Yes, and only the devfreq patches are required for that. # cat /sys/class/devfreq/devfreq0/available_frequencies 25000000 100000000 800000000 # cat /sys/class/devfreq/devfreq0/cur_freq 800000000 You should be able to control frequencies manually with the userspace governor: # echo "userspace" > /sys/class/devfreq/devfreq0/governor # echo "25000000" > /sys/class/devfreq/devfreq0/userspace/set_freq This series allows devices to request guaranteed bandwidth for themselves through the interconnect subsystem, without it DRAM freq will still switch but you'll have problems like display corruption as it turns on before freq goes up. You can check that probe worked by doing # cat /sys/kernel/debug/interconnect/interconnect_summary -- Regards, Leonard
Hi Leonard, On 9/16/19 05:34, Leonard Crestez wrote: > On 23.08.2019 17:37, Leonard Crestez wrote: >> This series add imx support for interconnect via devfreq: the ICC >> framework is used to aggregate requests from devices and then those are >> converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. >> >> Since there is no single devicetree node that can represent the "interconnect" >> new API is added to allow individual devfreq nodes to act as parsing proxies >> all mapping to a single soc-level icc provider. This is still RFC >> because of this > > Any comments? I made a lot of changes relative to previous versions, > most of them solely to avoid adding a virtual node in DT bindings. > > The only current interconnect provider implementation is for qcom and it > uses a firmware node as the provider node (with #interconnect-cells). > However there is no obvious equivalent of that for imx and many other SOCs. Not sure if it will help, but have you seen the qcs404 interconnect driver? There is also mt8183 interconnect provider driver on LKML. > On imx there are multiple pieces of scalable fabric which can be defined > in DT as devfreq devices and it sort of makes sense to add > #interconnect-cells to those. However when it comes to describing the > SOC interconnect graph it's much more convenient to have a single > per-SOC platform driver. Is all the NoC configuration done only by ATF? Are there any NoC related memory mapped registers? Thanks, Georgi
On 25.09.2019 05:38, Georgi Djakov wrote: > Hi Leonard, > > On 9/16/19 05:34, Leonard Crestez wrote: >> On 23.08.2019 17:37, Leonard Crestez wrote: >>> This series add imx support for interconnect via devfreq: the ICC >>> framework is used to aggregate requests from devices and then those are >>> converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. >>> >>> Since there is no single devicetree node that can represent the "interconnect" >>> new API is added to allow individual devfreq nodes to act as parsing proxies >>> all mapping to a single soc-level icc provider. This is still RFC >>> because of this >> >> Any comments? I made a lot of changes relative to previous versions, >> most of them solely to avoid adding a virtual node in DT bindings. >> >> The only current interconnect provider implementation is for qcom and it >> uses a firmware node as the provider node (with #interconnect-cells). >> However there is no obvious equivalent of that for imx and many other SOCs. > > Not sure if it will help, but have you seen the qcs404 interconnect driver? > There is also mt8183 interconnect provider driver on LKML. Yes, but only yesterday. The qcs404 driver involves multiple DT devices so it seems closer to imx. As far as I understand from reading qcs404 source: * There is no struct device representing the entire graph. * There are multiple NOCs and each registers itself as a separate interconnect provider. * Each NOC registers multiple icc_nodes of various sorts: * Device masters and slaves * Some nodes representing NoC ports? * Multiple internal nodes * There is single per-SOC master list of QNOCs in the qcs404 driver. * The QNOCs can reference each other between multiple providers. * Each NOC registers an icc_provider and a subset of the graph. * The multiple NoC inside a chip are distinguished by compat strings. This seems strange, aren't they really different instantiations of the same IP with small config changes? This design is still quite odd, what would make sense to me is to register the "interconnect graph" once and then provide multiple "interconnect scalers" which handle the aggregated requests for certain specific nodes. >> On imx there are multiple pieces of scalable fabric which can be defined >> in DT as devfreq devices and it sort of makes sense to add >> #interconnect-cells to those. However when it comes to describing the >> SOC interconnect graph it's much more convenient to have a single >> per-SOC platform driver. > > Is all the NoC configuration done only by ATF? Are there any NoC related memory > mapped registers? Registers are memory-mapped and visible to the A-cores but should only be accessed through secure transactions. This means that configuration needs be done by ATF in EL3 (we don't support running linux in secure world on imx8m). There is no "remote processor" managing this on imx8m. On older imx6/7 chips we actually have two out-of-tree implementations of bus freq switching code: An older one in Linux (used when running in secure world) and a different one in optee for running Linux in non-secure world. NoC registers can be used to control some "transaction priority" bits but I don't want to expose that part right now. What determines bandwidth versus power consumption is the NoC clk rate and clocks are managed by Linux directly. DVFS on the RAM controller (DDRC) is also important. That component is only a bus slave and frequency switching requires a complex sequence inside ATF. -- Regards, Leonard
On 9/25/19 15:52, Leonard Crestez wrote: > On 25.09.2019 05:38, Georgi Djakov wrote: >> Hi Leonard, >> >> On 9/16/19 05:34, Leonard Crestez wrote: >>> On 23.08.2019 17:37, Leonard Crestez wrote: >>>> This series add imx support for interconnect via devfreq: the ICC >>>> framework is used to aggregate requests from devices and then those are >>>> converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. >>>> >>>> Since there is no single devicetree node that can represent the "interconnect" >>>> new API is added to allow individual devfreq nodes to act as parsing proxies >>>> all mapping to a single soc-level icc provider. This is still RFC >>>> because of this >>> >>> Any comments? I made a lot of changes relative to previous versions, >>> most of them solely to avoid adding a virtual node in DT bindings. >>> >>> The only current interconnect provider implementation is for qcom and it >>> uses a firmware node as the provider node (with #interconnect-cells). >>> However there is no obvious equivalent of that for imx and many other SOCs. >> >> Not sure if it will help, but have you seen the qcs404 interconnect driver? >> There is also mt8183 interconnect provider driver on LKML. > > Yes, but only yesterday. The qcs404 driver involves multiple DT devices > so it seems closer to imx. > > As far as I understand from reading qcs404 source: > > * There is no struct device representing the entire graph. > * There are multiple NOCs and each registers itself as a separate > interconnect provider. > * Each NOC registers multiple icc_nodes of various sorts: > * Device masters and slaves > * Some nodes representing NoC ports? Well, all nodes are representing ports. > * Multiple internal nodes > * There is single per-SOC master list of QNOCs in the qcs404 driver. > * The QNOCs can reference each other between multiple providers. > * Each NOC registers an icc_provider and a subset of the graph. > * The multiple NoC inside a chip are distinguished by compat strings. > This seems strange, aren't they really different instantiations of the > same IP with small config changes? No, they are different IPs - ahb, axi or custom based. > This design is still quite odd, what would make sense to me is to > register the "interconnect graph" once and then provide multiple > "interconnect scalers" which handle the aggregated requests for certain > specific nodes. > >>> On imx there are multiple pieces of scalable fabric which can be defined >>> in DT as devfreq devices and it sort of makes sense to add >>> #interconnect-cells to those. However when it comes to describing the >>> SOC interconnect graph it's much more convenient to have a single >>> per-SOC platform driver. >> >> Is all the NoC configuration done only by ATF? Are there any NoC related memory >> mapped registers? > > Registers are memory-mapped and visible to the A-cores but should only > be accessed through secure transactions. This means that configuration > needs be done by ATF in EL3 (we don't support running linux in secure > world on imx8m). There is no "remote processor" managing this on imx8m. Can we create some noc DT node with it's memory mapped address and make it an interconnect provider? Sounds to me like a more correct representation of the hardware? Other option would be to bless some PSCI DT node (for example) to be a provider. > > On older imx6/7 chips we actually have two out-of-tree implementations > of bus freq switching code: An older one in Linux (used when running in > secure world) and a different one in optee for running Linux in > non-secure world. > > NoC registers can be used to control some "transaction priority" bits > but I don't want to expose that part right now. This is very similar to some of the Qcom hardware. > What determines bandwidth versus power consumption is the NoC clk rate > and clocks are managed by Linux directly. So you will need to describe these clocks in the interconnect provider DT node like on qcs404. > DVFS on the RAM controller (DDRC) is also important. That component is > only a bus slave and frequency switching requires a complex sequence > inside ATF. Makes sense. Thanks, Georgi
On 2019-09-26 3:52 AM, Georgi Djakov wrote: > On 9/25/19 15:52, Leonard Crestez wrote: >> On 25.09.2019 05:38, Georgi Djakov wrote: >>> Hi Leonard, >>> >>> On 9/16/19 05:34, Leonard Crestez wrote: >>>> On 23.08.2019 17:37, Leonard Crestez wrote: >>>>> This series add imx support for interconnect via devfreq: the ICC >>>>> framework is used to aggregate requests from devices and then those are >>>>> converted to DEV_PM_QOS_MIN_FREQUENCY requests for devfreq. >>>>> >>>>> Since there is no single devicetree node that can represent the "interconnect" >>>>> new API is added to allow individual devfreq nodes to act as parsing proxies >>>>> all mapping to a single soc-level icc provider. This is still RFC >>>>> because of this >>>> >>>> Any comments? I made a lot of changes relative to previous versions, >>>> most of them solely to avoid adding a virtual node in DT bindings. >>>> >>>> The only current interconnect provider implementation is for qcom and it >>>> uses a firmware node as the provider node (with #interconnect-cells). >>>> However there is no obvious equivalent of that for imx and many other SOCs. >>> >>> Not sure if it will help, but have you seen the qcs404 interconnect driver? >>> There is also mt8183 interconnect provider driver on LKML. >> >> Yes, but only yesterday. The qcs404 driver involves multiple DT devices >> so it seems closer to imx. >> >> As far as I understand from reading qcs404 source: >> >> * There is no struct device representing the entire graph. >> * There are multiple NOCs and each registers itself as a separate >> interconnect provider. >> * Each NOC registers multiple icc_nodes of various sorts: >> * Device masters and slaves >> * Some nodes representing NoC ports? > > Well, all nodes are representing ports. > >> * Multiple internal nodes >> * There is single per-SOC master list of QNOCs in the qcs404 driver. >> * The QNOCs can reference each other between multiple providers. >> * Each NOC registers an icc_provider and a subset of the graph. >> * The multiple NoC inside a chip are distinguished by compat strings. >> This seems strange, aren't they really different instantiations of the >> same IP with small config changes? > > No, they are different IPs - ahb, axi or custom based. On IMX some of the buses are just different instantiations. Would it make sense to standardize an "interconnect-node-id" to identify middle nodes? For example if you have nearly identical "audio" "display" "vpu" NICs then this property would make it possible to map from a DT done to an ICC graph node. >> This design is still quite odd, what would make sense to me is to >> register the "interconnect graph" once and then provide multiple >> "interconnect scalers" which handle the aggregated requests for certain >> specific nodes. >> >>>> On imx there are multiple pieces of scalable fabric which can be defined >>>> in DT as devfreq devices and it sort of makes sense to add >>>> #interconnect-cells to those. However when it comes to describing the >>>> SOC interconnect graph it's much more convenient to have a single >>>> per-SOC platform driver. >>> >>> Is all the NoC configuration done only by ATF? Are there any NoC related memory >>> mapped registers? >> >> Registers are memory-mapped and visible to the A-cores but should only >> be accessed through secure transactions. This means that configuration >> needs be done by ATF in EL3 (we don't support running linux in secure >> world on imx8m). There is no "remote processor" managing this on imx8m. > > Can we create some noc DT node with it's memory mapped address and make > it an interconnect provider? Sounds to me like a more correct representation > of the hardware? This is what I did, it's just that the initial binding is in this series: https://patchwork.kernel.org/cover/11104113/ https://patchwork.kernel.org/patch/11104137/ https://patchwork.kernel.org/patch/11104141/ The nodes are scaled via devfreq and interconnect comes "on top" to make device bandwidth requests. I think using devfreq is valuable for example: * DDRC can support reactive scaling based on performance counters * The NOC can run at different voltages so it should have it's own OPP table. > Other option would be to bless some PSCI DT node (for example) to be a > provider. I don't think this can be a good fit, I want to support different interconnect nodes with different underlying interfaces on the same SOC. There is no abstraction layer in firmware so abstractions for different interconnect midnodes should be in linux instead. >> On older imx6/7 chips we actually have two out-of-tree implementations >> of bus freq switching code: An older one in Linux (used when running in >> secure world) and a different one in optee for running Linux in >> non-secure world. >> >> NoC registers can be used to control some "transaction priority" bits >> but I don't want to expose that part right now. > > This is very similar to some of the Qcom hardware. NoC IP is licensed from Arteris which was bought-out by Qcom. Documentation is not public though and there are likely many differences versus what Qcom has in their own chips. >> What determines bandwidth versus power consumption is the NoC clk rate >> and clocks are managed by Linux directly. > > So you will need to describe these clocks in the interconnect provider > DT node like on qcs404. I already implemented the nodes as devfreq provider and DDRC even includes ondemand reactive scaling support: https://patchwork.kernel.org/patch/11104139/ https://patchwork.kernel.org/patch/11104145/ https://patchwork.kernel.org/patch/11104143/ I could just pick the "main" NOC and turn than into the "only" interconnect provider. Something like this: if (has_property(noc_device, "#interconnect-cells")) { register_soc_icc_driver(noc_device); } This would get rid of the icc_proxy stuff but fetching references to other scalable nodes would require some other way to identify them. -- Regards, Leonard