| Message ID | 20221020215318.4193269-1-opendmb@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | mm: introduce Designated Movable Blocks | expand |
On Thu, Oct 20, 2022 at 02:53:09PM -0700, Doug Berger wrote: > MOTIVATION: > Some Broadcom devices (e.g. 7445, 7278) contain multiple memory > controllers with each mapped in a different address range within > a Uniform Memory Architecture. Some users of these systems have > expressed the desire to locate ZONE_MOVABLE memory on each > memory controller to allow user space intensive processing to > make better use of the additional memory bandwidth. > Unfortunately, the historical monotonic layout of zones would > mean that if the lowest addressed memory controller contains > ZONE_MOVABLE memory then all of the memory available from > memory controllers at higher addresses must also be in the > ZONE_MOVABLE zone. This would force all kernel memory accesses > onto the lowest addressed memory controller and significantly > reduce the amount of memory available for non-movable > allocations. > I didn't review the first version of this patch because others, particularly David Hildenbrand highlighted many of the concerns I had. I broadly followed the discussion but didn't respond because I live in a permanent state of having too much to do but with a new version, I have to say something. The three big questions he initially asked were How large are these areas typically? How large are they in comparison to other memory in the system? How is this memory currently presented to the system? Can you share some more how exactly ZONE_MOVABLE would help here to make better use of the memory bandwidth? Zones are about addressing limitations primarily and frankly, ZONE_MOVABLE was a bad idea in retrospect. Today, the preferred approach would have been to create a separate NUMA node with distance-1 to the local node (fudge by adding 1 to the local distance "10" for zonelist purposes) that was ZONE_MOVABLE with the zonelists structured such that GFP_MOVABLE allocations would prefer the "movable" node first. While I don't recall why I did not take that approach, it most likely was because CONFIG_NUMA was not always set, it was only intended for hugetlbfs allocations and maybe I didn't have the necessary skill or foresight to take that approach. Hotplugs requirements are somewhat different, the primary motivation that I'm aware of is being able to guarantee they can be offlined, particularly nodes, which can be done in some circumstances. Generally hotplug does not care what uses the memory as long as it can be removed later. The requirements for restricted access to high speed memory is different. There is a high degree of uncertainity of how these regions are to be used by applications to get access to the high speed memory, to quote I'm not certain what is typical because these systems are highly configurable and Broadcom's customers have different ideas about application processing. ... The Designated Movable Block concept introduced here has the potential to offer useful services to different constituencies. I tried to highlight this in my V1 patch set with the hope of attracting some interest, but it can complicate the overall discussion, so I would like to maybe narrow the discussion here. It may be good to keep them in mind when assessing the overall value, but perhaps the "other opportunities" can be covered as a follow on discussion. I note the "potential" part here because we don't actually know. A major limitation of ZONE_MOVABLE is that there is no way of controlling access from userspace to restrict the high-speed memory to a designated application, only to all applications in general. The primary interface to control access to memory with different characteristics is mempolicies which is NUMA orientated, not zone orientated. So, if there is a special application that requires exclusive access, it's very difficult to configure based on zones. Furthermore, page table pages mapping data located in the high-speed region are stored in the slower memory which potentially impacts the performance if the working set of the application exceeds TLB reach. Finally, while there is mention that Broadcom may have some special interface to determine what applications can use the high-speed region, it's hardware-specific as opposed to something that belongs in the core mm. I agree that keeping the high-speed memory in a local node and using "sticky" pageblocks or CMA has limitations of its own but in itself, that does not justify using ZONE_MOVABLE in my opinion. The statement that ARM can have multiple controllers with equal distance and bandwidth (if I'm reading it correctly) but places them in different zones.... that's just a bit weird if there are no other addressing limitations. It's not obvious why ARM would do that, but it also does not matter because it shouldn't be a core mm concern. There are already examples of where memory is physically "local" to the CPU but has different bandwidth or latency including High Bandwidth (HBM), Sub-NUMA Clustering (SNC), PMEM as a memory-life device and some AMD EPYC Chips, particularly the first generation where a sockets memory controllers had different distances. With the broadcom controllers, it sounds like a local memory controller but the bandwidth available differs. It's functionally equivalent to HBM. The fact that the memory access is physically local to the CPU socket is irrelevant when the characteristics of that locality differs. NUMA stands for Non-Uniform Memory Access and if bandwidth to different address ranges differs, then the system is inherently NUMA even if that is inconvenient. While I have not evaluated the implementation in detail, there is already infrastructure dealing with tiered memory (memory that is local but has different characteristics) with support for moving memory between tiers depending on access patterns. Memory policies can be used to restrict access to what processes can access the higher bandwidth memory. Given the use case for DSM, I suspect that the intent is "application data uses high bandwidth memory where possible and kernel uses lower bandwidth memory" which is probably fine for an appliance because there is only one workload but it's not a generic solution suitable. Going back to the original questions; How large are these areas typically? How large are they in comparison to other memory in the system? I am treating this as the same question because the consequencs are the same. A high ratio of !MOVABLE:MOVABLE can cause big problems including premature OOM, surprising reclaim behaviour etc How is this memory currently presented to the system? It's local, but with different characteristics so it's inherently NUMA because it's Non-Uniform, there is no getting away from that. Can you share some more how exactly ZONE_MOVABLE would help here to make better use of the memory bandwidth? In the appliance case, it doesn't matter if the intent is that "all application data should use high bandwidth memory where possible and the application phase behaviour is predictable" and that may very well work fine for the users of the Broadcom platforms with multiple memory controllers. It does not work at all for the general where access must be restricted to a subset of tasks in a general system that can only be controlled with memory policies. The high bandwidth memory should be representated as a NUMA node, optionally to create that node as ZONE_MOVABLE and relying on the zonelists to select the movable zone as the first preference.
On 26.10.22 12:55, Mel Gorman wrote: > On Thu, Oct 20, 2022 at 02:53:09PM -0700, Doug Berger wrote: >> MOTIVATION: >> Some Broadcom devices (e.g. 7445, 7278) contain multiple memory >> controllers with each mapped in a different address range within >> a Uniform Memory Architecture. Some users of these systems have >> expressed the desire to locate ZONE_MOVABLE memory on each >> memory controller to allow user space intensive processing to >> make better use of the additional memory bandwidth. >> Unfortunately, the historical monotonic layout of zones would >> mean that if the lowest addressed memory controller contains >> ZONE_MOVABLE memory then all of the memory available from >> memory controllers at higher addresses must also be in the >> ZONE_MOVABLE zone. This would force all kernel memory accesses >> onto the lowest addressed memory controller and significantly >> reduce the amount of memory available for non-movable >> allocations. >> > > I didn't review the first version of this patch because others, particularly > David Hildenbrand highlighted many of the concerns I had. I broadly followed > the discussion but didn't respond because I live in a permanent state of > having too much to do but with a new version, I have to say something. :) Just a note that I am still behind on replying to the discussion in v2. I wish I had more capacity right now to be more responsive -- but just like you (Mel) "permanent state of having too much to do". Other things (especially bug fixes) have higher priority. Thanks for having a look at it Mel --- I only skimmed over your reply, but ... > > The three big questions he initially asked were > > How large are these areas typically? > How large are they in comparison to other memory in the system? > How is this memory currently presented to the system? > Can you share some more how exactly ZONE_MOVABLE would help here to make > better use of the memory bandwidth? > > Zones are about addressing limitations primarily and frankly, ZONE_MOVABLE > was a bad idea in retrospect. Today, the preferred approach would have > been to create a separate NUMA node with distance-1 to the local node > (fudge by adding 1 to the local distance "10" for zonelist purposes) > that was ZONE_MOVABLE with the zonelists structured such that GFP_MOVABLE > allocations would prefer the "movable" node first. While I don't recall > why I did not take that approach, it most likely was because CONFIG_NUMA > was not always set, it was only intended for hugetlbfs allocations and > maybe I didn't have the necessary skill or foresight to take that approach. > > Hotplugs requirements are somewhat different, the primary motivation that > I'm aware of is being able to guarantee they can be offlined, particularly > nodes, which can be done in some circumstances. Generally hotplug does > not care what uses the memory as long as it can be removed later. The > requirements for restricted access to high speed memory is different. > > There is a high degree of uncertainity of how these regions are to be > used by applications to get access to the high speed memory, to quote > > I'm not certain what is typical because these systems are highly > configurable and Broadcom's customers have different ideas about > application processing. > > ... > > The Designated Movable Block concept introduced here has the > potential to offer useful services to different constituencies. I > tried to highlight this in my V1 patch set with the hope of > attracting some interest, but it can complicate the overall > discussion, so I would like to maybe narrow the discussion here. It > may be good to keep them in mind when assessing the overall value, > but perhaps the "other opportunities" can be covered as a follow > on discussion. > > I note the "potential" part here because we don't actually know. A > major limitation of ZONE_MOVABLE is that there is no way of controlling > access from userspace to restrict the high-speed memory to a designated > application, only to all applications in general. The primary interface > to control access to memory with different characteristics is mempolicies > which is NUMA orientated, not zone orientated. So, if there is a special > application that requires exclusive access, it's very difficult to configure > based on zones. Furthermore, page table pages mapping data located in the > high-speed region are stored in the slower memory which potentially impacts > the performance if the working set of the application exceeds TLB reach. > Finally, while there is mention that Broadcom may have some special > interface to determine what applications can use the high-speed region, > it's hardware-specific as opposed to something that belongs in the core mm. > > I agree that keeping the high-speed memory in a local node and using "sticky" > pageblocks or CMA has limitations of its own but in itself, that does not > justify using ZONE_MOVABLE in my opinion. The statement that ARM can have > multiple controllers with equal distance and bandwidth (if I'm reading it > correctly) but places them in different zones.... that's just a bit weird if > there are no other addressing limitations. It's not obvious why ARM would do > that, but it also does not matter because it shouldn't be a core mm concern. > > There are already examples of where memory is physically "local" to > the CPU but has different bandwidth or latency including High Bandwidth > (HBM), Sub-NUMA Clustering (SNC), PMEM as a memory-life device and some > AMD EPYC Chips, particularly the first generation where a sockets memory > controllers had different distances. With the broadcom controllers, > it sounds like a local memory controller but the bandwidth available > differs. It's functionally equivalent to HBM. > > The fact that the memory access is physically local to the CPU socket is > irrelevant when the characteristics of that locality differs. NUMA stands > for Non-Uniform Memory Access and if bandwidth to different address ranges > differs, then the system is inherently NUMA even if that is inconvenient. > > While I have not evaluated the implementation in detail, there is already > infrastructure dealing with tiered memory (memory that is local but has > different characteristics) with support for moving memory between tiers > depending on access patterns. Memory policies can be used to restrict > access to what processes can access the higher bandwidth memory. Given the > use case for DSM, I suspect that the intent is "application data uses high > bandwidth memory where possible and kernel uses lower bandwidth memory" > which is probably fine for an appliance because there is only one workload > but it's not a generic solution suitable. > > Going back to the original questions; > > How large are these areas typically? > How large are they in comparison to other memory in the system? > > I am treating this as the same question because the consequencs are the > same. A high ratio of !MOVABLE:MOVABLE can cause big problems including > premature OOM, surprising reclaim behaviour etc > > How is this memory currently presented to the system? > > It's local, but with different characteristics so it's inherently NUMA > because it's Non-Uniform, there is no getting away from that. > > Can you share some more how exactly ZONE_MOVABLE would help here to make > better use of the memory bandwidth? > > In the appliance case, it doesn't matter if the intent is that "all > application data should use high bandwidth memory where possible and > the application phase behaviour is predictable" and that may very well > work fine for the users of the Broadcom platforms with multiple memory > controllers. It does not work at all for the general where access must > be restricted to a subset of tasks in a general system that can only be > controlled with memory policies. > > The high bandwidth memory should be representated as a NUMA node, optionally > to create that node as ZONE_MOVABLE and relying on the zonelists to select > the movable zone as the first preference. ... that boils down to my remark to tiered memory and eventually using devdax to expose this memory to the system and letting the admin decide to online it to ZONE_MOVABLE. Of course, that's just one way of doing it.
On Wed, Oct 26, 2022 at 01:11:40PM +0200, David Hildenbrand wrote: > > In the appliance case, it doesn't matter if the intent is that "all > > application data should use high bandwidth memory where possible and > > the application phase behaviour is predictable" and that may very well > > work fine for the users of the Broadcom platforms with multiple memory > > controllers. It does not work at all for the general where access must > > be restricted to a subset of tasks in a general system that can only be > > controlled with memory policies. > > > > The high bandwidth memory should be representated as a NUMA node, optionally > > to create that node as ZONE_MOVABLE and relying on the zonelists to select > > the movable zone as the first preference. > > ... that boils down to my remark to tiered memory and eventually using > devdax to expose this memory to the system and letting the admin decide to > online it to ZONE_MOVABLE. Of course, that's just one way of doing it. > I don't see this approach being inherently bad as such, particularly in the appliance space where it is known in advance what exactly is running and what the requirements are. It's not automagical but it's not worse than specifying something like movablecore=100M@2G,100M@3G,1G@1024G. In either case, knowledge of the address ranges needing special treatment is required with the difference being that access to the special memory can be restricted by policies in the general case.
On 10/26/2022 3:55 AM, Mel Gorman wrote: > On Thu, Oct 20, 2022 at 02:53:09PM -0700, Doug Berger wrote: >> MOTIVATION: >> Some Broadcom devices (e.g. 7445, 7278) contain multiple memory >> controllers with each mapped in a different address range within >> a Uniform Memory Architecture. Some users of these systems have >> expressed the desire to locate ZONE_MOVABLE memory on each >> memory controller to allow user space intensive processing to >> make better use of the additional memory bandwidth. >> Unfortunately, the historical monotonic layout of zones would >> mean that if the lowest addressed memory controller contains >> ZONE_MOVABLE memory then all of the memory available from >> memory controllers at higher addresses must also be in the >> ZONE_MOVABLE zone. This would force all kernel memory accesses >> onto the lowest addressed memory controller and significantly >> reduce the amount of memory available for non-movable >> allocations. >> > > I didn't review the first version of this patch because others, particularly > David Hildenbrand highlighted many of the concerns I had. I broadly followed > the discussion but didn't respond because I live in a permanent state of > having too much to do but with a new version, I have to say something. I am familiar with that state and as a beneficiary of your hard work I'll take the opportunity to say thanks and I appreciate you taking the time to respond. > > The three big questions he initially asked were > > How large are these areas typically? > How large are they in comparison to other memory in the system? > How is this memory currently presented to the system? > Can you share some more how exactly ZONE_MOVABLE would help here to make > better use of the memory bandwidth? > > Zones are about addressing limitations primarily and frankly, ZONE_MOVABLE > was a bad idea in retrospect. Today, the preferred approach would have > been to create a separate NUMA node with distance-1 to the local node > (fudge by adding 1 to the local distance "10" for zonelist purposes) > that was ZONE_MOVABLE with the zonelists structured such that GFP_MOVABLE > allocations would prefer the "movable" node first. I'm afraid I don't completely follow what you are suggesting here. > While I don't recall > why I did not take that approach, it most likely was because CONFIG_NUMA > was not always set, it was only intended for hugetlbfs allocations and > maybe I didn't have the necessary skill or foresight to take that approach. It remains true that CONFIG_NUMA is not always set and that is a key motivator for this patch set. For example, Google is moving to a common GKI kernel for their Google TV platform that they are requiring vendors to support. Currently the arm64 GKI kernel does not set CONFIG_NUMA and it seems unlikely that we will be able to get all vendors to accept such a change. > > Hotplugs requirements are somewhat different, the primary motivation that > I'm aware of is being able to guarantee they can be offlined, particularly > nodes, which can be done in some circumstances. Generally hotplug does > not care what uses the memory as long as it can be removed later. The > requirements for restricted access to high speed memory is different. This is effectively the same requirement that an implementation of 'reusable' reserved memory has. A driver that owns reserved memory may not care what uses the memory as long as the memory can be reclaimed when the driver needs it. This is functionally analogous to memory hotplug. Reserved memory that is 'reusable' and compatible with 'shared-dma-pool' uses the CMA implementation, but there is room for an alternative implementation that shares the memory more aggressively. This is a separate motivator for Designated Movable Block support, but I am deferring that discussion since it is desirable to have a more extended debate over APIs and such. > > There is a high degree of uncertainity of how these regions are to be > used by applications to get access to the high speed memory, to quote > > I'm not certain what is typical because these systems are highly > configurable and Broadcom's customers have different ideas about > application processing. > > ... > > The Designated Movable Block concept introduced here has the > potential to offer useful services to different constituencies. I > tried to highlight this in my V1 patch set with the hope of > attracting some interest, but it can complicate the overall > discussion, so I would like to maybe narrow the discussion here. It > may be good to keep them in mind when assessing the overall value, > but perhaps the "other opportunities" can be covered as a follow > on discussion. > > I note the "potential" part here because we don't actually know. I used "potential" here not as in "it might be useful", but rather that "different constituencies (i.e. people outside of the Broadcom ecosystem) might also find them useful". > A > major limitation of ZONE_MOVABLE is that there is no way of controlling > access from userspace to restrict the high-speed memory to a designated > application, only to all applications in general. The primary interface > to control access to memory with different characteristics is mempolicies > which is NUMA orientated, not zone orientated. So, if there is a special > application that requires exclusive access, it's very difficult to configure > based on zones. Furthermore, page table pages mapping data located in the > high-speed region are stored in the slower memory which potentially impacts > the performance if the working set of the application exceeds TLB reach. > Finally, while there is mention that Broadcom may have some special > interface to determine what applications can use the high-speed region, > it's hardware-specific as opposed to something that belongs in the core mm. > > I agree that keeping the high-speed memory in a local node and using "sticky" > pageblocks or CMA has limitations of its own but in itself, that does not > justify using ZONE_MOVABLE in my opinion. The statement that ARM can have > multiple controllers with equal distance and bandwidth (if I'm reading it > correctly) but places them in different zones.... that's just a bit weird if > there are no other addressing limitations. It's not obvious why ARM would do > that, but it also does not matter because it shouldn't be a core mm concern. There appears to be some confusion regarding my explanation of multiple memory controllers on a device like the BCM7278. There is no inherent performance difference between the two memory controllers and their attached DRAM. They merely provide the opportunity to perform memory accesses in parallel for different physical address ranges. The physical address ranges were selected by the SoC designers for reasons only known to them, but I'm sure they had no consideration of zones in their decision making. The selection of zones remains an artifact of the design of Linux. Since the BCM7278 contains a 4-core SMP cluster and each core can have multiple outstanding memory transactions the speed of DDR transactions can create a bottleneck for the system. If each memory controller has an effective bandwidth of X then, provided the system memory accesses can be distributed across both memory controllers, the combined effective bandwidth can be additive (X + X = 2X). Of course the actual result is highly dependent on the dependent clause "provided the system memory accesses can be distributed across both memory controllers". The accesses do not need to be evenly distributed to gain a benefit. We just want to reduce any idle time on each memory controller. It was observed that the monotonic zone layout for a non-NUMA system (like this one) creates a bias for kernel space to use lower physical memory addresses and user space to use higher physical memory addresses. Broadcom customers requested the ability to locate movablecore memory within the physical address range of each memory controller and reported that it improved their system performance. Unfortunately, I do not have access to their data and I doubt they would allow me to share it if I did. I don't believe this is really about trying to optimize the performance of a specific application as much as trying to prevent overall system performance degradation from underutilized memory bandwidth. > > There are already examples of where memory is physically "local" to > the CPU but has different bandwidth or latency including High Bandwidth > (HBM), Sub-NUMA Clustering (SNC), PMEM as a memory-life device and some > AMD EPYC Chips, particularly the first generation where a sockets memory > controllers had different distances. With the broadcom controllers, > it sounds like a local memory controller but the bandwidth available > differs. It's functionally equivalent to HBM. The bandwidth available does not differ, but if too few transactions target one of the memory controllers, that controllers bandwidth is underutilized. > > The fact that the memory access is physically local to the CPU socket is > irrelevant when the characteristics of that locality differs. NUMA stands > for Non-Uniform Memory Access and if bandwidth to different address ranges > differs, then the system is inherently NUMA even if that is inconvenient. The bandwidth to different address ranges does not differ. A single threaded application should see no performance difference regardless of where its memory is located. However, if multiple application threads are executing in parallel and the memory is divided between the memory controllers they will be able to do more work per unit of time than if the memory is predominantly located on one memory controller. > > While I have not evaluated the implementation in detail, there is already > infrastructure dealing with tiered memory (memory that is local but has > different characteristics) with support for moving memory between tiers > depending on access patterns. Memory policies can be used to restrict > access to what processes can access the higher bandwidth memory. Given the > use case for DSM, I suspect that the intent is "application data uses high > bandwidth memory where possible and kernel uses lower bandwidth memory" > which is probably fine for an appliance because there is only one workload > but it's not a generic solution suitable. > > Going back to the original questions; > > How large are these areas typically? > How large are they in comparison to other memory in the system? > > I am treating this as the same question because the consequencs are the > same. A high ratio of !MOVABLE:MOVABLE can cause big problems including > premature OOM, surprising reclaim behaviour etc This is what makes the current movablecore implementation unacceptable. In order to get any movablecore memory in the lower physical address range requires all of the upper physical address range to be movablecore which is horribly unbalanced. Specifying a value like 'movablecore=256M@2G,512M' with this patch set allows us to specify 512MB of total movablecore with 256MB of it at a physical address within the lower memory controller and the remainder at the highest addresses of the upper memory controller. > > How is this memory currently presented to the system? > > It's local, but with different characteristics so it's inherently NUMA > because it's Non-Uniform, there is no getting away from that. It does not have different characteristics. > > Can you share some more how exactly ZONE_MOVABLE would help here to make > better use of the memory bandwidth? > > In the appliance case, it doesn't matter if the intent is that "all > application data should use high bandwidth memory where possible and > the application phase behaviour is predictable" and that may very well > work fine for the users of the Broadcom platforms with multiple memory > controllers. It does not work at all for the general where access must > be restricted to a subset of tasks in a general system that can only be > controlled with memory policies. > > The high bandwidth memory should be representated as a NUMA node, optionally > to create that node as ZONE_MOVABLE and relying on the zonelists to select > the movable zone as the first preference. > This patch set is fundamentally about greater control over the placement of movablecore memory. The current implementation of movablecore requires all of the ZONE_MOVABLE memory to be located at the highest physical addresses of the system when CONFIG_NUMA is not set. Allowing the specification of a base address allows greater flexibility on systems where there are benefits. Thanks again for your time, Doug
On Wed, Nov 02, 2022 at 03:33:53PM -0700, Doug Berger wrote: > > > > The three big questions he initially asked were > > > > How large are these areas typically? > > How large are they in comparison to other memory in the system? > > How is this memory currently presented to the system? > > Can you share some more how exactly ZONE_MOVABLE would help here to make > > better use of the memory bandwidth? > > > > Zones are about addressing limitations primarily and frankly, ZONE_MOVABLE > > was a bad idea in retrospect. Today, the preferred approach would have > > been to create a separate NUMA node with distance-1 to the local node > > (fudge by adding 1 to the local distance "10" for zonelist purposes) > > that was ZONE_MOVABLE with the zonelists structured such that GFP_MOVABLE > > allocations would prefer the "movable" node first. > > I'm afraid I don't completely follow what you are suggesting here. > It's not especially important how it could have been done but using a node would have avoided confusing zones (address limitations) with memory partitioning (e.g. MOVABLE). Nodes can also interleave but it would have required CONFIG_NUMA so pointless for GKI and the current discussion other than with a time machine, GKI might have enabled CONFIG_NUMA :/ > > While I don't recall > > why I did not take that approach, it most likely was because CONFIG_NUMA > > was not always set, it was only intended for hugetlbfs allocations and > > maybe I didn't have the necessary skill or foresight to take that approach. > > It remains true that CONFIG_NUMA is not always set and that is a key > motivator for this patch set. For example, Google is moving to a common GKI > kernel for their Google TV platform that they are requiring vendors to > support. Currently the arm64 GKI kernel does not set CONFIG_NUMA and it > seems unlikely that we will be able to get all vendors to accept such a > change. > Ok. > > > > Hotplugs requirements are somewhat different, the primary motivation that > > I'm aware of is being able to guarantee they can be offlined, particularly > > nodes, which can be done in some circumstances. Generally hotplug does > > not care what uses the memory as long as it can be removed later. The > > requirements for restricted access to high speed memory is different. > > This is effectively the same requirement that an implementation of > 'reusable' reserved memory has. A driver that owns reserved memory may not > care what uses the memory as long as the memory can be reclaimed when the > driver needs it. This is functionally analogous to memory hotplug. Reserved > memory that is 'reusable' and compatible with 'shared-dma-pool' uses the CMA > implementation, but there is room for an alternative implementation that > shares the memory more aggressively. This is a separate motivator for > Designated Movable Block support, but I am deferring that discussion since > it is desirable to have a more extended debate over APIs and such. > There needs to be a better explanation as to why CMA cannot be used or more importantly why page_alloc.shuffle= should not be used (more on that later. It's not clear how a movable zone shares memory more aggressively than CMA would. Both have the problem that the if protected range is too large that premature memory exhaustion can occur for kernel allocations. > > > > There is a high degree of uncertainity of how these regions are to be > > used by applications to get access to the high speed memory, to quote > > > > I'm not certain what is typical because these systems are highly > > configurable and Broadcom's customers have different ideas about > > application processing. > > > > ... > > > > The Designated Movable Block concept introduced here has the > > potential to offer useful services to different constituencies. I > > tried to highlight this in my V1 patch set with the hope of > > attracting some interest, but it can complicate the overall > > discussion, so I would like to maybe narrow the discussion here. It > > may be good to keep them in mind when assessing the overall value, > > but perhaps the "other opportunities" can be covered as a follow > > on discussion. > > > > I note the "potential" part here because we don't actually know. > > I used "potential" here not as in "it might be useful", but rather that > "different constituencies (i.e. people outside of the Broadcom ecosystem) > might also find them useful". > That's very vague unfortunately. > > A > > major limitation of ZONE_MOVABLE is that there is no way of controlling > > access from userspace to restrict the high-speed memory to a designated > > application, only to all applications in general. The primary interface > > to control access to memory with different characteristics is mempolicies > > which is NUMA orientated, not zone orientated. So, if there is a special > > application that requires exclusive access, it's very difficult to configure > > based on zones. Furthermore, page table pages mapping data located in the > > high-speed region are stored in the slower memory which potentially impacts > > the performance if the working set of the application exceeds TLB reach. > > Finally, while there is mention that Broadcom may have some special > > interface to determine what applications can use the high-speed region, > > it's hardware-specific as opposed to something that belongs in the core mm. > > > > I agree that keeping the high-speed memory in a local node and using "sticky" > > pageblocks or CMA has limitations of its own but in itself, that does not > > justify using ZONE_MOVABLE in my opinion. The statement that ARM can have > > multiple controllers with equal distance and bandwidth (if I'm reading it > > correctly) but places them in different zones.... that's just a bit weird if > > there are no other addressing limitations. It's not obvious why ARM would do > > that, but it also does not matter because it shouldn't be a core mm concern. > > There appears to be some confusion regarding my explanation of multiple > memory controllers on a device like the BCM7278. There is no inherent > performance difference between the two memory controllers and their attached > DRAM. They merely provide the opportunity to perform memory accesses in > parallel for different physical address ranges. The physical address ranges > were selected by the SoC designers for reasons only known to them, but I'm > sure they had no consideration of zones in their decision making. The > selection of zones remains an artifact of the design of Linux. > Ok, so the channels are equal but the channels are not interleaved in hardware so basically you are trying to implement software-based memory channel interleaving? > Since the BCM7278 contains a 4-core SMP cluster and each core can have > multiple outstanding memory transactions the speed of DDR transactions can > create a bottleneck for the system. If each memory controller has an > effective bandwidth of X then, provided the system memory accesses can be > distributed across both memory controllers, the combined effective bandwidth > can be additive (X + X = 2X). Of course the actual result is highly > dependent on the dependent clause "provided the system memory accesses can > be distributed across both memory controllers". The accesses do not need to > be evenly distributed to gain a benefit. We just want to reduce any idle > time on each memory controller. > > It was observed that the monotonic zone layout for a non-NUMA system (like > this one) creates a bias for kernel space to use lower physical memory > addresses and user space to use higher physical memory addresses. Broadcom > customers requested the ability to locate movablecore memory within the > physical address range of each memory controller and reported that it > improved their system performance. Unfortunately, I do not have access to > their data and I doubt they would allow me to share it if I did. I don't > believe this is really about trying to optimize the performance of a > specific application as much as trying to prevent overall system performance > degradation from underutilized memory bandwidth. > So if I'm reading this right, the intent behind using ZONE_MOVABLE at fixed address ranges is so most (but not all) user pages end up using one controller and all kernel pages and some user pages use the other controller. If kernel and userspace accesses are split 50/50, then the memory bandwidth usage will be split across channels. However, if the ratio of kernel:user accesses is large then the bandwidth usage will still be assymetric. For example, if there are 10 times more accesses to kernel pages then user pages, then one channel will receive most of the traffic. The reverse is also true but to a lesser extent as user pages can use all zones and kernel accesses use a subset. Depending on the access pattern, creating separate zones may not help at all and in some cases, could make the problem worse. The trap is that it might happen to work for a fixed appliance like a TV with a predictable workload, it may not work in the general case. Splitting based on the __GFP_MOVABLE does not guarantee that idle time on a memory controller can be reduced as it relies on the access pattern. > > There are already examples of where memory is physically "local" to > > the CPU but has different bandwidth or latency including High Bandwidth > > (HBM), Sub-NUMA Clustering (SNC), PMEM as a memory-life device and some > > AMD EPYC Chips, particularly the first generation where a sockets memory > > controllers had different distances. With the broadcom controllers, > > it sounds like a local memory controller but the bandwidth available > > differs. It's functionally equivalent to HBM. > > The bandwidth available does not differ, but if too few transactions target > one of the memory controllers, that controllers bandwidth is underutilized. > This is also a limitation of the patch series. Lets say the bulk of accesses are to user pages allocated in ZONE_MOVABLE which correlates to one memory channel then one channel gets saturated anyway. It also gets more complicated if there are more controllers because the only division possible is between MOVABLE/everything else. An odd number of channels will be difficult to split meaningfully. > > The fact that the memory access is physically local to the CPU socket is > > irrelevant when the characteristics of that locality differs. NUMA stands > > for Non-Uniform Memory Access and if bandwidth to different address ranges > > differs, then the system is inherently NUMA even if that is inconvenient. > > The bandwidth to different address ranges does not differ. A single threaded > application should see no performance difference regardless of where its > memory is located. However, if multiple application threads are executing in > parallel and the memory is divided between the memory controllers they will > be able to do more work per unit of time than if the memory is predominantly > located on one memory controller. > And if multiple application threads dominantly access user pages then splitting the zone will not necessarily help, particularly if ZONE_MOVABLE is not filled as the bulk of the accesses will still use one memory channel. > > In the appliance case, it doesn't matter if the intent is that "all > > application data should use high bandwidth memory where possible and > > the application phase behaviour is predictable" and that may very well > > work fine for the users of the Broadcom platforms with multiple memory > > controllers. It does not work at all for the general where access must > > be restricted to a subset of tasks in a general system that can only be > > controlled with memory policies. > > > > The high bandwidth memory should be representated as a NUMA node, optionally > > to create that node as ZONE_MOVABLE and relying on the zonelists to select > > the movable zone as the first preference. > > > This patch set is fundamentally about greater control over the placement of > movablecore memory. The current implementation of movablecore requires all > of the ZONE_MOVABLE memory to be located at the highest physical addresses > of the system when CONFIG_NUMA is not set. Allowing the specification of a > base address allows greater flexibility on systems where there are benefits. > Unfortunately, while greater control of the ranges used by ZONE_MOVABLE will help in some cases, it will not help in others and may be misleading. If memory accesses need to be interleaved in software then the free lists need to be broken up into arenas within a zone by some boundary whether that boundary is is fixed-length ranges, memory sections, memory channels or specified explicitly on the kernel command line. Any allocation type can use any arena with tasks moving to another arena based on contention, pageblock type availability or interleaving round-robin explicitly. Unfortunately, it's non-trivial to implement and a *lot* of heavy lifting. A somewhat awful hack would be to reorder top-level MAX_ORDER-1 list at boot time. By default that list is ordered 1, 2, 3 ...... n-2, n-1, n If at boot time it was reordered to be 1, n, 2, n-1, 3, n-2 ...... This would interleave all the early allocations across memory channels in the case where channels are based on large contiguous physical ranges of memory. Applications starting early would then interleave between channels but after a period of time, it would be pseudo-random and it's weak. A similar, and probably better, option is to look at what page_alloc.shuffle= does and randomly shuffle the free lists to randomly select between the memory channels. I haven't looked at the implementation recently and forget how it works exactly. Maybe it would benefit from being able to take ranges that should be special cased for shuffling, particularly at boot time to order it "1, n, 2, n-1" as described above or allowing SHUFFLE_ORDER to be a lower value. Either way, shuffling would achieve similar goals of spreading allocations between channels without assuming that the access ratio of kernel:user is close to 1:1.
On 11/18/2022 9:05 AM, Mel Gorman wrote: > On Wed, Nov 02, 2022 at 03:33:53PM -0700, Doug Berger wrote: >>> >>> Hotplugs requirements are somewhat different, the primary motivation that >>> I'm aware of is being able to guarantee they can be offlined, particularly >>> nodes, which can be done in some circumstances. Generally hotplug does >>> not care what uses the memory as long as it can be removed later. The >>> requirements for restricted access to high speed memory is different. >> >> This is effectively the same requirement that an implementation of >> 'reusable' reserved memory has. A driver that owns reserved memory may not >> care what uses the memory as long as the memory can be reclaimed when the >> driver needs it. This is functionally analogous to memory hotplug. Reserved >> memory that is 'reusable' and compatible with 'shared-dma-pool' uses the CMA >> implementation, but there is room for an alternative implementation that >> shares the memory more aggressively. This is a separate motivator for >> Designated Movable Block support, but I am deferring that discussion since >> it is desirable to have a more extended debate over APIs and such. >> > > There needs to be a better explanation as to why CMA cannot be used or more > importantly why page_alloc.shuffle= should not be used (more on that later. > It's not clear how a movable zone shares memory more aggressively than CMA > would. Both have the problem that the if protected range is too large that > premature memory exhaustion can occur for kernel allocations. The pages within a CMA pool can be allocated by either the CMA allocator (i.e. alloc_contig_range()) for use by the kernel (i.e. drivers) or by the page allocator to satisfy movable requests (i.e. predominantly user space). However, the page allocator is constrained by rules that make allocations from MIGRATE_CMA free lists a secondary consideration when the memory is available elsewhere. This tends toward keeping pages in a CMA pool free while user processes consume more memory outside of the CMA pool that could have been used more generally by the kernel. Pages on the MIGRATE_MOVABLE free list in ZONE_MOVABLE (i.e. all of the pages in the movable zone) are the first choice for satisfying movable requests. This allows user space to make full use of a 'reusable' reserved memory range that isn't actively used by a driver, which is what I mean by "more aggressive". When the driver wants to reclaim its 'reusable' reserved memory range it can use alloc_contig_range() to force any movable allocations out of the range perhaps into memory that could have been used more generally by the kernel. Such a reclamation may be more time consuming to complete since the pages are more likely to be in use than if they were on the MIGRATE_CMA free list, but no pages go unused by either the driver or user space. I was not familiar with page_alloc.shuffle, but it may very well have a role to play here. > >>> >>> There is a high degree of uncertainity of how these regions are to be >>> used by applications to get access to the high speed memory, to quote >>> >>> I'm not certain what is typical because these systems are highly >>> configurable and Broadcom's customers have different ideas about >>> application processing. >>> >>> ... >>> >>> The Designated Movable Block concept introduced here has the >>> potential to offer useful services to different constituencies. I >>> tried to highlight this in my V1 patch set with the hope of >>> attracting some interest, but it can complicate the overall >>> discussion, so I would like to maybe narrow the discussion here. It >>> may be good to keep them in mind when assessing the overall value, >>> but perhaps the "other opportunities" can be covered as a follow >>> on discussion. >>> >>> I note the "potential" part here because we don't actually know. >> >> I used "potential" here not as in "it might be useful", but rather that >> "different constituencies (i.e. people outside of the Broadcom ecosystem) >> might also find them useful". >> > > That's very vague unfortunately. As an example, there have been submissions related to memory hotplug that could have used this capability if it existed at the time. In fact the comments for [1] incorrectly assumed the movablecore= behavior proposed here was already implemented. Eventually those patch sets morphed into the movable_node implementation and modification of movablecore= was dropped. The point I intended to make is that in addition to the use case identified here (i.e. improved bandwidth utilization on multiple memory controller systems) Designated Movable Blocks as a general mechanism could be useful to people I don't know to solve problems of which I am not aware (e.g. memory hot unplugging, reusable reserved memory, ???). I offer as evidence that I am not the only person to conceive of the concept. [1] https://lore.kernel.org/all/1374220774-29974-21-git-send-email-tangchen@cn.fujitsu.com/ > >>> A >>> major limitation of ZONE_MOVABLE is that there is no way of controlling >>> access from userspace to restrict the high-speed memory to a designated >>> application, only to all applications in general. The primary interface >>> to control access to memory with different characteristics is mempolicies >>> which is NUMA orientated, not zone orientated. So, if there is a special >>> application that requires exclusive access, it's very difficult to configure >>> based on zones. Furthermore, page table pages mapping data located in the >>> high-speed region are stored in the slower memory which potentially impacts >>> the performance if the working set of the application exceeds TLB reach. >>> Finally, while there is mention that Broadcom may have some special >>> interface to determine what applications can use the high-speed region, >>> it's hardware-specific as opposed to something that belongs in the core mm. >>> >>> I agree that keeping the high-speed memory in a local node and using "sticky" >>> pageblocks or CMA has limitations of its own but in itself, that does not >>> justify using ZONE_MOVABLE in my opinion. The statement that ARM can have >>> multiple controllers with equal distance and bandwidth (if I'm reading it >>> correctly) but places them in different zones.... that's just a bit weird if >>> there are no other addressing limitations. It's not obvious why ARM would do >>> that, but it also does not matter because it shouldn't be a core mm concern. >> >> There appears to be some confusion regarding my explanation of multiple >> memory controllers on a device like the BCM7278. There is no inherent >> performance difference between the two memory controllers and their attached >> DRAM. They merely provide the opportunity to perform memory accesses in >> parallel for different physical address ranges. The physical address ranges >> were selected by the SoC designers for reasons only known to them, but I'm >> sure they had no consideration of zones in their decision making. The >> selection of zones remains an artifact of the design of Linux. >> > > Ok, so the channels are equal but the channels are not interleaved in > hardware so basically you are trying to implement software-based memory > channel interleaving? I suppose that could be a fair characterization of the objective, though the approach taken here is very much a "poor man's" approach that attempts to improve things without requiring the "heavy lifting" required for a more complete solution. That said, the use of page_alloc.shuffle is interesting here since it could improve the likelyhood of interleaving. > >> Since the BCM7278 contains a 4-core SMP cluster and each core can have >> multiple outstanding memory transactions the speed of DDR transactions can >> create a bottleneck for the system. If each memory controller has an >> effective bandwidth of X then, provided the system memory accesses can be >> distributed across both memory controllers, the combined effective bandwidth >> can be additive (X + X = 2X). Of course the actual result is highly >> dependent on the dependent clause "provided the system memory accesses can >> be distributed across both memory controllers". The accesses do not need to >> be evenly distributed to gain a benefit. We just want to reduce any idle >> time on each memory controller. >> >> It was observed that the monotonic zone layout for a non-NUMA system (like >> this one) creates a bias for kernel space to use lower physical memory >> addresses and user space to use higher physical memory addresses. Broadcom >> customers requested the ability to locate movablecore memory within the >> physical address range of each memory controller and reported that it >> improved their system performance. Unfortunately, I do not have access to >> their data and I doubt they would allow me to share it if I did. I don't >> believe this is really about trying to optimize the performance of a >> specific application as much as trying to prevent overall system performance >> degradation from underutilized memory bandwidth. >> > > So if I'm reading this right, the intent behind using ZONE_MOVABLE at > fixed address ranges is so most (but not all) user pages end up using > one controller and all kernel pages and some user pages use the other > controller. If kernel and userspace accesses are split 50/50, then the > memory bandwidth usage will be split across channels. However, if the > ratio of kernel:user accesses is large then the bandwidth usage will still > be assymetric. > > For example, if there are 10 times more accesses to kernel pages then user > pages, then one channel will receive most of the traffic. The reverse > is also true but to a lesser extent as user pages can use all zones and > kernel accesses use a subset. Depending on the access pattern, creating > separate zones may not help at all and in some cases, could make the problem > worse. The trap is that it might happen to work for a fixed appliance like > a TV with a predictable workload, it may not work in the general case. > > Splitting based on the __GFP_MOVABLE does not guarantee that idle time on > a memory controller can be reduced as it relies on the access pattern. You are not reading it quite right. We could accomplish the split you describe on a BCM7278 SoC with two memory controllers using the existing movablecore=50% kernel parameter. This would create a ZONE_MOVABLE on the high address memory controller and a ZONE_DMA on the low address memory controller. What is of interest to Broadcom customers is to better distribute user space accesses across each memory controller to improve the bandwidth available to user space dominated work flows. With no ZONE_MOVABLE, the BCM7278 SoC with 1GB of memory on each memory controller will place the 1GB on the low address memory controller in ZONE_DMA and the 1GB on the high address memory controller in ZONE_NORMAL. With this layout movable allocation requests will only fallback to the ZONE_DMA (low memory controller) once the ZONE_NORMAL (high memory controller) is sufficiently depleted of free memory. Adding ZONE_MOVABLE memory above ZONE_NORMAL with the current movablecore behavior does not improve this situation other than forcing more kernel allocations off of the high memory controller. User space allocations are even more likely to be on the high memory controller. The Designated Movable Block mechanism allows ZONE_MOVABLE memory to be located on the low memory controller to make it easier for user space allocations to land on the low memory controller. If ZONE_MOVABLE is only placed on the low memory controller then user space allocations can land in ZONE_NORMAL on the high memory controller, but only through fallback after ZONE_MOVABLE is sufficiently depleted of free memory which is just the reverse of the existing situation. The Designated Movable Block mechanism allows ZONE_MOVABLE memory to be located on each memory controller so that user space allocations have equal access to each memory controller until the ZONE_MOVABLE memory is depleted and fallback to other zones occurs. To my knowledge Broadcom customers that are currently using the Designated Movable Block mechanism are relying on the somewhat random starting and stopping of parallel user space processes to produce a more random distribution of ZONE_MOVABLE allocations across multiple memory controllers, but the page_alloc.shuffle mechanism seems like it would be a good addition to promote this randomness. Even better, it appears that page_alloc.shuffle is already enabled in the GKI configuration. You are of course correct that the access patterns make all of the difference and it is almost certain that one memory controller or the other will be saturated at any given time, but the intent is to increase the opportunity to use more of the total bandwidth made available by the multiple memory controllers. > >>> There are already examples of where memory is physically "local" to >>> the CPU but has different bandwidth or latency including High Bandwidth >>> (HBM), Sub-NUMA Clustering (SNC), PMEM as a memory-life device and some >>> AMD EPYC Chips, particularly the first generation where a sockets memory >>> controllers had different distances. With the broadcom controllers, >>> it sounds like a local memory controller but the bandwidth available >>> differs. It's functionally equivalent to HBM. >> >> The bandwidth available does not differ, but if too few transactions target >> one of the memory controllers, that controllers bandwidth is underutilized. >> > > This is also a limitation of the patch series. Lets say the bulk of > accesses are to user pages allocated in ZONE_MOVABLE which correlates to > one memory channel then one channel gets saturated anyway. > > It also gets more complicated if there are more controllers because the > only division possible is between MOVABLE/everything else. An odd number > of channels will be difficult to split meaningfully. The patch series allows Designated Movable Blocks to occupy a portion of each memory controller while allowing the ZONE_MOVABLE zone to span all of the memory controllers. In this way user pages allocated from ZONE_MOVABLE may be distributed across all of the memory controllers. Use of page_alloc.shuffle should improve the randomness of this distribution. Memory outside of Designated Movable Blocks on each memory controller can be outside ZONE_MOVABLE (e.g. ZONE_DMA and ZONE_NORMAL) and managed accordingly. An odd number of channels need not affect this. > >>> The fact that the memory access is physically local to the CPU socket is >>> irrelevant when the characteristics of that locality differs. NUMA stands >>> for Non-Uniform Memory Access and if bandwidth to different address ranges >>> differs, then the system is inherently NUMA even if that is inconvenient. >> >> The bandwidth to different address ranges does not differ. A single threaded >> application should see no performance difference regardless of where its >> memory is located. However, if multiple application threads are executing in >> parallel and the memory is divided between the memory controllers they will >> be able to do more work per unit of time than if the memory is predominantly >> located on one memory controller. >> > > And if multiple application threads dominantly access user pages then > splitting the zone will not necessarily help, particularly if ZONE_MOVABLE > is not filled as the bulk of the accesses will still use one memory channel. > >>> In the appliance case, it doesn't matter if the intent is that "all >>> application data should use high bandwidth memory where possible and >>> the application phase behaviour is predictable" and that may very well >>> work fine for the users of the Broadcom platforms with multiple memory >>> controllers. It does not work at all for the general where access must >>> be restricted to a subset of tasks in a general system that can only be >>> controlled with memory policies. >>> >>> The high bandwidth memory should be representated as a NUMA node, optionally >>> to create that node as ZONE_MOVABLE and relying on the zonelists to select >>> the movable zone as the first preference. >>> >> This patch set is fundamentally about greater control over the placement of >> movablecore memory. The current implementation of movablecore requires all >> of the ZONE_MOVABLE memory to be located at the highest physical addresses >> of the system when CONFIG_NUMA is not set. Allowing the specification of a >> base address allows greater flexibility on systems where there are benefits. >> > > Unfortunately, while greater control of the ranges used by ZONE_MOVABLE > will help in some cases, it will not help in others and may be misleading. > > If memory accesses need to be interleaved in software then the free lists > need to be broken up into arenas within a zone by some boundary whether > that boundary is is fixed-length ranges, memory sections, memory channels > or specified explicitly on the kernel command line. Any allocation type > can use any arena with tasks moving to another arena based on contention, > pageblock type availability or interleaving round-robin explicitly. > Unfortunately, it's non-trivial to implement and a *lot* of heavy lifting. > > A somewhat awful hack would be to reorder top-level MAX_ORDER-1 list at > boot time. By default that list is ordered > > 1, 2, 3 ...... n-2, n-1, n > > If at boot time it was reordered to be > > 1, n, 2, n-1, 3, n-2 ...... > > This would interleave all the early allocations across memory channels in > the case where channels are based on large contiguous physical ranges of > memory. Applications starting early would then interleave between channels > but after a period of time, it would be pseudo-random and it's weak. > > A similar, and probably better, option is to look at what page_alloc.shuffle= > does and randomly shuffle the free lists to randomly select between the > memory channels. I haven't looked at the implementation recently and forget > how it works exactly. Maybe it would benefit from being able to take ranges > that should be special cased for shuffling, particularly at boot time to > order it "1, n, 2, n-1" as described above or allowing SHUFFLE_ORDER to > be a lower value. Either way, shuffling would achieve similar goals of > spreading allocations between channels without assuming that the access > ratio of kernel:user is close to 1:1. > I decided to implement this very simple multi-threaded application as a testcase to experiment with the concepts discussed here: #include <stdio.h> #include <stdlib.h> #include <string.h> #include <pthread.h> #define BUF_SIZE (0x4000000) #define THREADS (4) #define COPY_COUNT (30) void *thread_function( void *ptr ); int main() { pthread_t thread[THREADS]; int i, iret[THREADS]; for(i = 0; i < THREADS; i++) iret[i] = pthread_create( &thread[i], NULL, thread_function, (void*) NULL); for(i = 0; i < THREADS; i++) pthread_join( thread[i], NULL); for(i = 0; i < THREADS; i++) printf("Thread %d returns: %d\n", i, iret[i]); exit(0); } void *thread_function( void *ptr ) { char *s, *d; int i; s = malloc(BUF_SIZE); if (!s) return NULL; d = malloc(BUF_SIZE); if (!d) { free(s); return NULL; } for (i = 0; i < COPY_COUNT; i++) { memcpy(d, s, BUF_SIZE); } free(s); free(d); } It meaninglessly moves data from one large dynamically allocated buffer to another a number of times without trying to be clever. I experimented with a Broadcom BCM7278 system with 1GB on each memory controller (i.e. 2GB total memory). The buffers were made large to render data caching meaningless and to require several pages to be allocated to populate the buffer. With V3 of this patch set applied to a 6.1-rc1 kernel I observed these results: With no movablecore kernel parameter specified: # time /tmp/thread_test Thread 1 returns: 0 Thread 2 returns: 0 Thread 3 returns: 0 Thread 4 returns: 0 real 0m4.047s user 0m14.183s sys 0m1.215s With this additional kernel parameter "movablecore=600M": # time /tmp/thread_test Thread 0 returns: 0 Thread 1 returns: 0 Thread 2 returns: 0 Thread 3 returns: 0 real 0m4.068s user 0m14.402s sys 0m1.117s With this additional kernel parameter "movablecore=600M@0x50000000": # time /tmp/thread_test Thread 0 returns: 0 Thread 1 returns: 0 Thread 2 returns: 0 Thread 3 returns: 0 real 0m4.010s user 0m13.979s sys 0m1.070s However, with these additional kernel parameters "movablecore=300M@0x60000000,300M@0x320000000 page_alloc.shuffle=1": # time /tmp/thread_test Thread 0 returns: 0 Thread 1 returns: 0 Thread 2 returns: 0 Thread 3 returns: 0 real 0m3.173s user 0m11.175s sys 0m1.067s These numbers show an over 20% improvement in performance of the test application when distributing ZONE_MOVABLE across both memory controllers. Happy Holidays! Doug
On 10/20/22 14:53, Doug Berger wrote: > MOTIVATION: > Some Broadcom devices (e.g. 7445, 7278) contain multiple memory > controllers with each mapped in a different address range within > a Uniform Memory Architecture. Some users of these systems have > expressed the desire to locate ZONE_MOVABLE memory on each > memory controller to allow user space intensive processing to > make better use of the additional memory bandwidth. > Unfortunately, the historical monotonic layout of zones would > mean that if the lowest addressed memory controller contains > ZONE_MOVABLE memory then all of the memory available from > memory controllers at higher addresses must also be in the > ZONE_MOVABLE zone. This would force all kernel memory accesses > onto the lowest addressed memory controller and significantly > reduce the amount of memory available for non-movable > allocations. > > The main objective of this patch set is therefore to allow a > block of memory to be designated as part of the ZONE_MOVABLE > zone where it will always only be used by the kernel page > allocator to satisfy requests for movable pages. The term > Designated Movable Block is introduced here to represent such a > block. The favored implementation allows extension of the > 'movablecore' kernel parameter to allow specification of a base > address and support for multiple blocks. The existing > 'movablecore' mechanisms are retained. > > BACKGROUND: > NUMA architectures support distributing movablecore memory > across each node, but it is undesirable to introduce the > overhead and complexities of NUMA on systems that don't have a > Non-Uniform Memory Architecture. > > Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror option") > also depends on zone overlap to support sytems with multiple > mirrored ranges. > > Commit c6f03e2903c9 ("mm, memory_hotplug: remove zone restrictions") > embraced overlapped zones for memory hotplug. > > This commit set follows their lead to allow the ZONE_MOVABLE > zone to overlap other zones. Designated Movable Blocks are made > absent from overlapping zones and present within the > ZONE_MOVABLE zone. > > I initially investigated an implementation using a Designated > Movable migrate type in line with comments[1] made by Mel Gorman > regarding a "sticky" MIGRATE_MOVABLE type to avoid using > ZONE_MOVABLE. However, this approach was riskier since it was > much more instrusive on the allocation paths. Ultimately, the > progress made by the memory hotplug folks to expand the > ZONE_MOVABLE functionality convinced me to follow this approach. > Mel, David, does the sub-thread discussion with Doug help ensuring that all of the context is gathered before getting into a more detailed patch review on a patch-by-patch basis? Eventually we may need a fairly firm answer as to whether the proposed approach has any chance of landing upstream in order to either commit to in subsequent iterations of this patch set, or find an alternative. Thank you!
On Wed, Dec 14, 2022 at 04:17:35PM -0800, Doug Berger wrote: > On 11/18/2022 9:05 AM, Mel Gorman wrote: > > On Wed, Nov 02, 2022 at 03:33:53PM -0700, Doug Berger wrote: > > > > > > > > Hotplugs requirements are somewhat different, the primary motivation that > > > > I'm aware of is being able to guarantee they can be offlined, particularly > > > > nodes, which can be done in some circumstances. Generally hotplug does > > > > not care what uses the memory as long as it can be removed later. The > > > > requirements for restricted access to high speed memory is different. > > > > > > This is effectively the same requirement that an implementation of > > > 'reusable' reserved memory has. A driver that owns reserved memory may not > > > care what uses the memory as long as the memory can be reclaimed when the > > > driver needs it. This is functionally analogous to memory hotplug. Reserved > > > memory that is 'reusable' and compatible with 'shared-dma-pool' uses the CMA > > > implementation, but there is room for an alternative implementation that > > > shares the memory more aggressively. This is a separate motivator for > > > Designated Movable Block support, but I am deferring that discussion since > > > it is desirable to have a more extended debate over APIs and such. > > > > > > > There needs to be a better explanation as to why CMA cannot be used or more > > importantly why page_alloc.shuffle= should not be used (more on that later. > > It's not clear how a movable zone shares memory more aggressively than CMA > > would. Both have the problem that the if protected range is too large that > > premature memory exhaustion can occur for kernel allocations. > > The pages within a CMA pool can be allocated by either the CMA allocator > (i.e. alloc_contig_range()) for use by the kernel (i.e. drivers) or by the > page allocator to satisfy movable requests (i.e. predominantly user space). > However, the page allocator is constrained by rules that make allocations > from MIGRATE_CMA free lists a secondary consideration when the memory is > available elsewhere. This tends toward keeping pages in a CMA pool free > while user processes consume more memory outside of the CMA pool that could > have been used more generally by the kernel. > Pages on the MIGRATE_MOVABLE free list in ZONE_MOVABLE (i.e. all of the > pages in the movable zone) are the first choice for satisfying movable > requests. This allows user space to make full use of a 'reusable' reserved > memory range that isn't actively used by a driver, which is what I mean by > "more aggressive". When the driver wants to reclaim its 'reusable' reserved > memory range it can use alloc_contig_range() to force any movable > allocations out of the range perhaps into memory that could have been used > more generally by the kernel. Such a reclamation may be more time consuming > to complete since the pages are more likely to be in use than if they were > on the MIGRATE_CMA free list, but no pages go unused by either the driver or > user space. > Ok, so CMA is a bad fit. It's off the table. > I was not familiar with page_alloc.shuffle, but it may very well have a role > to play here. > It almost certainly does because unlike zones or CMA, it affects how free lists are arranged. IIRC, the original purpose was about improving performance of high-speed direct-mapped cache but it also serves a purpose in this case -- randomising allocations between two channels. It's still not perfect interleaving but better than none. > > > > A > > > > major limitation of ZONE_MOVABLE is that there is no way of controlling > > > > access from userspace to restrict the high-speed memory to a designated > > > > application, only to all applications in general. The primary interface > > > > to control access to memory with different characteristics is mempolicies > > > > which is NUMA orientated, not zone orientated. So, if there is a special > > > > application that requires exclusive access, it's very difficult to configure > > > > based on zones. Furthermore, page table pages mapping data located in the > > > > high-speed region are stored in the slower memory which potentially impacts > > > > the performance if the working set of the application exceeds TLB reach. > > > > Finally, while there is mention that Broadcom may have some special > > > > interface to determine what applications can use the high-speed region, > > > > it's hardware-specific as opposed to something that belongs in the core mm. > > > > > > > > I agree that keeping the high-speed memory in a local node and using "sticky" > > > > pageblocks or CMA has limitations of its own but in itself, that does not > > > > justify using ZONE_MOVABLE in my opinion. The statement that ARM can have > > > > multiple controllers with equal distance and bandwidth (if I'm reading it > > > > correctly) but places them in different zones.... that's just a bit weird if > > > > there are no other addressing limitations. It's not obvious why ARM would do > > > > that, but it also does not matter because it shouldn't be a core mm concern. > > > > > > There appears to be some confusion regarding my explanation of multiple > > > memory controllers on a device like the BCM7278. There is no inherent > > > performance difference between the two memory controllers and their attached > > > DRAM. They merely provide the opportunity to perform memory accesses in > > > parallel for different physical address ranges. The physical address ranges > > > were selected by the SoC designers for reasons only known to them, but I'm > > > sure they had no consideration of zones in their decision making. The > > > selection of zones remains an artifact of the design of Linux. > > > > > > > Ok, so the channels are equal but the channels are not interleaved in > > hardware so basically you are trying to implement software-based memory > > channel interleaving? > > I suppose that could be a fair characterization of the objective, though the > approach taken here is very much a "poor man's" approach that attempts to > improve things without requiring the "heavy lifting" required for a more > complete solution. > It's still unfortunate that the concept of zones being primarily about addressing or capability limitations changes. It's also difficult to use as any user of it has to be very aware of the memory channel configuration of the machine and know how to match addresses to channels. Information from zoneinfo on start_pfns, spanned ranges and the like become less useful. It's relatively minor but splitting the zones also means there is a performance hit during compaction because pageblock_pfn_to_page is more expensive. > That said, the use of page_alloc.shuffle is interesting here since it could > improve the likelyhood of interleaving. > Exactly. > > > > > Since the BCM7278 contains a 4-core SMP cluster and each core can have > > > multiple outstanding memory transactions the speed of DDR transactions can > > > create a bottleneck for the system. If each memory controller has an > > > effective bandwidth of X then, provided the system memory accesses can be > > > distributed across both memory controllers, the combined effective bandwidth > > > can be additive (X + X = 2X). Of course the actual result is highly > > > dependent on the dependent clause "provided the system memory accesses can > > > be distributed across both memory controllers". The accesses do not need to > > > be evenly distributed to gain a benefit. We just want to reduce any idle > > > time on each memory controller. > > > > > > It was observed that the monotonic zone layout for a non-NUMA system (like > > > this one) creates a bias for kernel space to use lower physical memory > > > addresses and user space to use higher physical memory addresses. Broadcom > > > customers requested the ability to locate movablecore memory within the > > > physical address range of each memory controller and reported that it > > > improved their system performance. Unfortunately, I do not have access to > > > their data and I doubt they would allow me to share it if I did. I don't > > > believe this is really about trying to optimize the performance of a > > > specific application as much as trying to prevent overall system performance > > > degradation from underutilized memory bandwidth. > > > > > > > So if I'm reading this right, the intent behind using ZONE_MOVABLE at > > fixed address ranges is so most (but not all) user pages end up using > > one controller and all kernel pages and some user pages use the other > > controller. If kernel and userspace accesses are split 50/50, then the > > memory bandwidth usage will be split across channels. However, if the > > ratio of kernel:user accesses is large then the bandwidth usage will still > > be assymetric. > > > > For example, if there are 10 times more accesses to kernel pages then user > > pages, then one channel will receive most of the traffic. The reverse > > is also true but to a lesser extent as user pages can use all zones and > > kernel accesses use a subset. Depending on the access pattern, creating > > separate zones may not help at all and in some cases, could make the problem > > worse. The trap is that it might happen to work for a fixed appliance like > > a TV with a predictable workload, it may not work in the general case. > > > > Splitting based on the __GFP_MOVABLE does not guarantee that idle time on > > a memory controller can be reduced as it relies on the access pattern. > > You are not reading it quite right. We could accomplish the split you > describe on a BCM7278 SoC with two memory controllers using the existing > movablecore=50% kernel parameter. This would create a ZONE_MOVABLE on the > high address memory controller and a ZONE_DMA on the low address memory > controller. > Ok, I did misunderstand at the time that ZONE_MOVABLE would be split between the controllers to improve interleaving of user accesses. > What is of interest to Broadcom customers is to better distribute user space > accesses across each memory controller to improve the bandwidth available to > user space dominated work flows. With no ZONE_MOVABLE, the BCM7278 SoC with > 1GB of memory on each memory controller will place the 1GB on the low > address memory controller in ZONE_DMA and the 1GB on the high address memory > controller in ZONE_NORMAL. With this layout movable allocation requests will > only fallback to the ZONE_DMA (low memory controller) once the ZONE_NORMAL > (high memory controller) is sufficiently depleted of free memory. > > Adding ZONE_MOVABLE memory above ZONE_NORMAL with the current movablecore > behavior does not improve this situation other than forcing more kernel > allocations off of the high memory controller. User space allocations are > even more likely to be on the high memory controller. > But it's a weak promise that interleaving will happen. If only a portion of ZONE_MOVABLE is used, it might still be all on the same channel. This might improve over time if enough memory was used and the system was up for long enough. > The Designated Movable Block mechanism allows ZONE_MOVABLE memory to be > located on the low memory controller to make it easier for user space > allocations to land on the low memory controller. If ZONE_MOVABLE is only > placed on the low memory controller then user space allocations can land in > ZONE_NORMAL on the high memory controller, but only through fallback after > ZONE_MOVABLE is sufficiently depleted of free memory which is just the > reverse of the existing situation. The Designated Movable Block mechanism > allows ZONE_MOVABLE memory to be located on each memory controller so that > user space allocations have equal access to each memory controller until the > ZONE_MOVABLE memory is depleted and fallback to other zones occurs. > > To my knowledge Broadcom customers that are currently using the Designated > Movable Block mechanism are relying on the somewhat random starting and > stopping of parallel user space processes to produce a more random > distribution of ZONE_MOVABLE allocations across multiple memory controllers, > but the page_alloc.shuffle mechanism seems like it would be a good addition > to promote this randomness. Even better, it appears that page_alloc.shuffle > is already enabled in the GKI configuration. > The "random starting and stopping of parallel user space processes" is required for the mechanism to work. It's arbitrary and unknown if the interleaving happens where as shuffle has an immediate, if random, impact. > You are of course correct that the access patterns make all of the > difference and it is almost certain that one memory controller or the other > will be saturated at any given time, but the intent is to increase the > opportunity to use more of the total bandwidth made available by the > multiple memory controllers. > And shuffle should also provide that opportunity except it's trivial to configure and only requires the user to know the memory channels are not interleaved. > > > > > > There are already examples of where memory is physically "local" to > > > > the CPU but has different bandwidth or latency including High Bandwidth > > > > (HBM), Sub-NUMA Clustering (SNC), PMEM as a memory-life device and some > > > > AMD EPYC Chips, particularly the first generation where a sockets memory > > > > controllers had different distances. With the broadcom controllers, > > > > it sounds like a local memory controller but the bandwidth available > > > > differs. It's functionally equivalent to HBM. > > > > > > The bandwidth available does not differ, but if too few transactions target > > > one of the memory controllers, that controllers bandwidth is underutilized. > > > > > > > This is also a limitation of the patch series. Lets say the bulk of > > accesses are to user pages allocated in ZONE_MOVABLE which correlates to > > one memory channel then one channel gets saturated anyway. > > > > It also gets more complicated if there are more controllers because the > > only division possible is between MOVABLE/everything else. An odd number > > of channels will be difficult to split meaningfully. > > The patch series allows Designated Movable Blocks to occupy a portion of > each memory controller while allowing the ZONE_MOVABLE zone to span all of > the memory controllers. In this way user pages allocated from ZONE_MOVABLE > may be distributed across all of the memory controllers. Use of > page_alloc.shuffle should improve the randomness of this distribution. > Memory outside of Designated Movable Blocks on each memory controller can be > outside ZONE_MOVABLE (e.g. ZONE_DMA and ZONE_NORMAL) and managed > accordingly. An odd number of channels need not affect this. > You're right, if ZONE_MOVABLE is split across the channels then the number of channels is less important. > > > > > > The fact that the memory access is physically local to the CPU socket is > > > > irrelevant when the characteristics of that locality differs. NUMA stands > > > > for Non-Uniform Memory Access and if bandwidth to different address ranges > > > > differs, then the system is inherently NUMA even if that is inconvenient. > > > > > > The bandwidth to different address ranges does not differ. A single threaded > > > application should see no performance difference regardless of where its > > > memory is located. However, if multiple application threads are executing in > > > parallel and the memory is divided between the memory controllers they will > > > be able to do more work per unit of time than if the memory is predominantly > > > located on one memory controller. > > > > > > > And if multiple application threads dominantly access user pages then > > splitting the zone will not necessarily help, particularly if ZONE_MOVABLE > > is not filled as the bulk of the accesses will still use one memory channel. > > > > > > In the appliance case, it doesn't matter if the intent is that "all > > > > application data should use high bandwidth memory where possible and > > > > the application phase behaviour is predictable" and that may very well > > > > work fine for the users of the Broadcom platforms with multiple memory > > > > controllers. It does not work at all for the general where access must > > > > be restricted to a subset of tasks in a general system that can only be > > > > controlled with memory policies. > > > > > > > > The high bandwidth memory should be representated as a NUMA node, optionally > > > > to create that node as ZONE_MOVABLE and relying on the zonelists to select > > > > the movable zone as the first preference. > > > > > > > This patch set is fundamentally about greater control over the placement of > > > movablecore memory. The current implementation of movablecore requires all > > > of the ZONE_MOVABLE memory to be located at the highest physical addresses > > > of the system when CONFIG_NUMA is not set. Allowing the specification of a > > > base address allows greater flexibility on systems where there are benefits. > > > > > > > Unfortunately, while greater control of the ranges used by ZONE_MOVABLE > > will help in some cases, it will not help in others and may be misleading. > > > > If memory accesses need to be interleaved in software then the free lists > > need to be broken up into arenas within a zone by some boundary whether > > that boundary is is fixed-length ranges, memory sections, memory channels > > or specified explicitly on the kernel command line. Any allocation type > > can use any arena with tasks moving to another arena based on contention, > > pageblock type availability or interleaving round-robin explicitly. > > Unfortunately, it's non-trivial to implement and a *lot* of heavy lifting. > > > > A somewhat awful hack would be to reorder top-level MAX_ORDER-1 list at > > boot time. By default that list is ordered > > > > 1, 2, 3 ...... n-2, n-1, n > > > > If at boot time it was reordered to be > > > > 1, n, 2, n-1, 3, n-2 ...... > > > > This would interleave all the early allocations across memory channels in > > the case where channels are based on large contiguous physical ranges of > > memory. Applications starting early would then interleave between channels > > but after a period of time, it would be pseudo-random and it's weak. > > > > A similar, and probably better, option is to look at what page_alloc.shuffle= > > does and randomly shuffle the free lists to randomly select between the > > memory channels. I haven't looked at the implementation recently and forget > > how it works exactly. Maybe it would benefit from being able to take ranges > > that should be special cased for shuffling, particularly at boot time to > > order it "1, n, 2, n-1" as described above or allowing SHUFFLE_ORDER to > > be a lower value. Either way, shuffling would achieve similar goals of > > spreading allocations between channels without assuming that the access > > ratio of kernel:user is close to 1:1. > > > > I decided to implement this very simple multi-threaded application as a > testcase to experiment with the concepts discussed here: > #include <stdio.h> > #include <stdlib.h> > #include <string.h> > #include <pthread.h> > > #define BUF_SIZE (0x4000000) > #define THREADS (4) > #define COPY_COUNT (30) > > void *thread_function( void *ptr ); > > int main() > { > pthread_t thread[THREADS]; > int i, iret[THREADS]; > > for(i = 0; i < THREADS; i++) > iret[i] = pthread_create( &thread[i], NULL, thread_function, (void*) > NULL); > > for(i = 0; i < THREADS; i++) > pthread_join( thread[i], NULL); > > for(i = 0; i < THREADS; i++) > printf("Thread %d returns: %d\n", i, iret[i]); > exit(0); > } > > void *thread_function( void *ptr ) > { > char *s, *d; > int i; > > s = malloc(BUF_SIZE); > if (!s) > return NULL; > > d = malloc(BUF_SIZE); > if (!d) { > free(s); > return NULL; > } > > for (i = 0; i < COPY_COUNT; i++) { > memcpy(d, s, BUF_SIZE); > } > free(s); > free(d); > } > Straight-forward, it's a meaningless load but relevant to this problem. The buffers are small enough that all threads would likely sit on the same controller with a vanilla kernel and compete for bandwidth. > It meaninglessly moves data from one large dynamically allocated buffer to > another a number of times without trying to be clever. I experimented with a > Broadcom BCM7278 system with 1GB on each memory controller (i.e. 2GB total > memory). The buffers were made large to render data caching meaningless and > to require several pages to be allocated to populate the buffer. > > With V3 of this patch set applied to a 6.1-rc1 kernel I observed these > results: > With no movablecore kernel parameter specified: > # time /tmp/thread_test > Thread 1 returns: 0 > Thread 2 returns: 0 > Thread 3 returns: 0 > Thread 4 returns: 0 > > real 0m4.047s > user 0m14.183s > sys 0m1.215s > > With this additional kernel parameter "movablecore=600M": > # time /tmp/thread_test > Thread 0 returns: 0 > Thread 1 returns: 0 > Thread 2 returns: 0 > Thread 3 returns: 0 > > real 0m4.068s > user 0m14.402s > sys 0m1.117s > > With this additional kernel parameter "movablecore=600M@0x50000000": > # time /tmp/thread_test > Thread 0 returns: 0 > Thread 1 returns: 0 > Thread 2 returns: 0 > Thread 3 returns: 0 > > real 0m4.010s > user 0m13.979s > sys 0m1.070s > > However, with these additional kernel parameters > "movablecore=300M@0x60000000,300M@0x320000000 page_alloc.shuffle=1": > # time /tmp/thread_test > Thread 0 returns: 0 > Thread 1 returns: 0 > Thread 2 returns: 0 > Thread 3 returns: 0 > > real 0m3.173s > user 0m11.175s > sys 0m1.067s > What were the results with just "movablecore=300M@0x60000000,300M@0x320000000" on its own and page_alloc.shuffle=1 on its own? For shuffle on its own, my expectations are that the results will be variable, sometimes good and sometimes bad, because it's at the mercy of the randomisation. It might be slightly improved if the initial top-level lists were ordered "1, n, 2, n-1, 3, n-2" optionally in __shuffle_zone or if shuffle_pick_tail was aware of the memory channels but a lot more work to implement.
On 04.01.23 00:43, Florian Fainelli wrote: > On 10/20/22 14:53, Doug Berger wrote: >> MOTIVATION: >> Some Broadcom devices (e.g. 7445, 7278) contain multiple memory >> controllers with each mapped in a different address range within >> a Uniform Memory Architecture. Some users of these systems have >> expressed the desire to locate ZONE_MOVABLE memory on each >> memory controller to allow user space intensive processing to >> make better use of the additional memory bandwidth. >> Unfortunately, the historical monotonic layout of zones would >> mean that if the lowest addressed memory controller contains >> ZONE_MOVABLE memory then all of the memory available from >> memory controllers at higher addresses must also be in the >> ZONE_MOVABLE zone. This would force all kernel memory accesses >> onto the lowest addressed memory controller and significantly >> reduce the amount of memory available for non-movable >> allocations. >> >> The main objective of this patch set is therefore to allow a >> block of memory to be designated as part of the ZONE_MOVABLE >> zone where it will always only be used by the kernel page >> allocator to satisfy requests for movable pages. The term >> Designated Movable Block is introduced here to represent such a >> block. The favored implementation allows extension of the >> 'movablecore' kernel parameter to allow specification of a base >> address and support for multiple blocks. The existing >> 'movablecore' mechanisms are retained. >> >> BACKGROUND: >> NUMA architectures support distributing movablecore memory >> across each node, but it is undesirable to introduce the >> overhead and complexities of NUMA on systems that don't have a >> Non-Uniform Memory Architecture. >> >> Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror option") >> also depends on zone overlap to support sytems with multiple >> mirrored ranges. >> >> Commit c6f03e2903c9 ("mm, memory_hotplug: remove zone restrictions") >> embraced overlapped zones for memory hotplug. >> >> This commit set follows their lead to allow the ZONE_MOVABLE >> zone to overlap other zones. Designated Movable Blocks are made >> absent from overlapping zones and present within the >> ZONE_MOVABLE zone. >> >> I initially investigated an implementation using a Designated >> Movable migrate type in line with comments[1] made by Mel Gorman >> regarding a "sticky" MIGRATE_MOVABLE type to avoid using >> ZONE_MOVABLE. However, this approach was riskier since it was >> much more instrusive on the allocation paths. Ultimately, the >> progress made by the memory hotplug folks to expand the >> ZONE_MOVABLE functionality convinced me to follow this approach. >> > > Mel, David, does the sub-thread discussion with Doug help ensuring that > all of the context is gathered before getting into a more detailed patch > review on a patch-by-patch basis? > > Eventually we may need a fairly firm answer as to whether the proposed > approach has any chance of landing upstream in order to either commit to > in subsequent iterations of this patch set, or find an alternative. As raised, I'd appreciate if less intrusive alternatives could be evaluated (e.g., fake NUMA nodes and being ablee to just use mbind(), moving such memory to ZONE_MOVABLE after boot via something like daxctl). I'm not convinced that these intrusive changes are worth it at this point. Further, some of the assumptions (ZONE_MOVABLE == user space) are not really future proof as I raised.
On 1/4/23 07:56, David Hildenbrand wrote: > On 04.01.23 00:43, Florian Fainelli wrote: >> On 10/20/22 14:53, Doug Berger wrote: >>> MOTIVATION: >>> Some Broadcom devices (e.g. 7445, 7278) contain multiple memory >>> controllers with each mapped in a different address range within >>> a Uniform Memory Architecture. Some users of these systems have >>> expressed the desire to locate ZONE_MOVABLE memory on each >>> memory controller to allow user space intensive processing to >>> make better use of the additional memory bandwidth. >>> Unfortunately, the historical monotonic layout of zones would >>> mean that if the lowest addressed memory controller contains >>> ZONE_MOVABLE memory then all of the memory available from >>> memory controllers at higher addresses must also be in the >>> ZONE_MOVABLE zone. This would force all kernel memory accesses >>> onto the lowest addressed memory controller and significantly >>> reduce the amount of memory available for non-movable >>> allocations. >>> >>> The main objective of this patch set is therefore to allow a >>> block of memory to be designated as part of the ZONE_MOVABLE >>> zone where it will always only be used by the kernel page >>> allocator to satisfy requests for movable pages. The term >>> Designated Movable Block is introduced here to represent such a >>> block. The favored implementation allows extension of the >>> 'movablecore' kernel parameter to allow specification of a base >>> address and support for multiple blocks. The existing >>> 'movablecore' mechanisms are retained. >>> >>> BACKGROUND: >>> NUMA architectures support distributing movablecore memory >>> across each node, but it is undesirable to introduce the >>> overhead and complexities of NUMA on systems that don't have a >>> Non-Uniform Memory Architecture. >>> >>> Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror >>> option") >>> also depends on zone overlap to support sytems with multiple >>> mirrored ranges. >>> >>> Commit c6f03e2903c9 ("mm, memory_hotplug: remove zone restrictions") >>> embraced overlapped zones for memory hotplug. >>> >>> This commit set follows their lead to allow the ZONE_MOVABLE >>> zone to overlap other zones. Designated Movable Blocks are made >>> absent from overlapping zones and present within the >>> ZONE_MOVABLE zone. >>> >>> I initially investigated an implementation using a Designated >>> Movable migrate type in line with comments[1] made by Mel Gorman >>> regarding a "sticky" MIGRATE_MOVABLE type to avoid using >>> ZONE_MOVABLE. However, this approach was riskier since it was >>> much more instrusive on the allocation paths. Ultimately, the >>> progress made by the memory hotplug folks to expand the >>> ZONE_MOVABLE functionality convinced me to follow this approach. >>> >> >> Mel, David, does the sub-thread discussion with Doug help ensuring that >> all of the context is gathered before getting into a more detailed patch >> review on a patch-by-patch basis? >> >> Eventually we may need a fairly firm answer as to whether the proposed >> approach has any chance of landing upstream in order to either commit to >> in subsequent iterations of this patch set, or find an alternative. > > > As raised, I'd appreciate if less intrusive alternatives could be > evaluated (e.g., fake NUMA nodes and being ablee to just use mbind(), > moving such memory to ZONE_MOVABLE after boot via something like daxctl). This is not an option with the environment we have to ultimately fit in which is Android TV utilizing the GKI kernel which does not enable NUMA and probably never will, and for similar reasons bringing a whole swath of user-space tools like daxctl may not be practical either, from both a logistical perspective (simply getting the tools built with bionic, accepted etc.) as well as system configuration perspective. > > I'm not convinced that these intrusive changes are worth it at this > point. Further, some of the assumptions (ZONE_MOVABLE == user space) are > not really future proof as I raised. I find this patch set reasonably small in contrast to a lot of other mm/ changes, what did you find intrusive specifically? AFAICT, there only assumption that is being made is that ZONE_MOVABLE contains memory that can be moved, but even if it did not in the future, there should hopefully be enough opportunities, given a large enough DMB region to service the allocation requests of its users. I will go back and read your comment to make sure I don't misunderstand it. Thanks
On 1/4/23 07:43, Mel Gorman wrote: [snip] >> What is of interest to Broadcom customers is to better distribute user space >> accesses across each memory controller to improve the bandwidth available to >> user space dominated work flows. With no ZONE_MOVABLE, the BCM7278 SoC with >> 1GB of memory on each memory controller will place the 1GB on the low >> address memory controller in ZONE_DMA and the 1GB on the high address memory >> controller in ZONE_NORMAL. With this layout movable allocation requests will >> only fallback to the ZONE_DMA (low memory controller) once the ZONE_NORMAL >> (high memory controller) is sufficiently depleted of free memory. >> >> Adding ZONE_MOVABLE memory above ZONE_NORMAL with the current movablecore >> behavior does not improve this situation other than forcing more kernel >> allocations off of the high memory controller. User space allocations are >> even more likely to be on the high memory controller. >> > > But it's a weak promise that interleaving will happen. If only a portion > of ZONE_MOVABLE is used, it might still be all on the same channel. This > might improve over time if enough memory was used and the system was up > for long enough. It is indeed a weak promise for user-space allocations out of ZONE_MOVABLE, however the other consumer of the DMB region is a kernel driver (typically a video decoder engine) which is directly tied to a specific memory controller/DMB region. For the kernel driver using the DMB region there is a hard guarantee from the kernel that it gets memory from a specific PFN range mapping directly to the desired memory controller and thus it is meeting the desired bandwidth allocation/deadlines/bursts etc. We care about both sides of the coin, though we acknowledge that "controlling" where user-space allocations are coming from such that they be steered towards a specific memory controller is a much harder task and so having some amount of non-determinism is acceptable here.
On 04.01.23 20:00, Florian Fainelli wrote: > On 1/4/23 07:56, David Hildenbrand wrote: >> On 04.01.23 00:43, Florian Fainelli wrote: >>> On 10/20/22 14:53, Doug Berger wrote: >>>> MOTIVATION: >>>> Some Broadcom devices (e.g. 7445, 7278) contain multiple memory >>>> controllers with each mapped in a different address range within >>>> a Uniform Memory Architecture. Some users of these systems have >>>> expressed the desire to locate ZONE_MOVABLE memory on each >>>> memory controller to allow user space intensive processing to >>>> make better use of the additional memory bandwidth. >>>> Unfortunately, the historical monotonic layout of zones would >>>> mean that if the lowest addressed memory controller contains >>>> ZONE_MOVABLE memory then all of the memory available from >>>> memory controllers at higher addresses must also be in the >>>> ZONE_MOVABLE zone. This would force all kernel memory accesses >>>> onto the lowest addressed memory controller and significantly >>>> reduce the amount of memory available for non-movable >>>> allocations. >>>> >>>> The main objective of this patch set is therefore to allow a >>>> block of memory to be designated as part of the ZONE_MOVABLE >>>> zone where it will always only be used by the kernel page >>>> allocator to satisfy requests for movable pages. The term >>>> Designated Movable Block is introduced here to represent such a >>>> block. The favored implementation allows extension of the >>>> 'movablecore' kernel parameter to allow specification of a base >>>> address and support for multiple blocks. The existing >>>> 'movablecore' mechanisms are retained. >>>> >>>> BACKGROUND: >>>> NUMA architectures support distributing movablecore memory >>>> across each node, but it is undesirable to introduce the >>>> overhead and complexities of NUMA on systems that don't have a >>>> Non-Uniform Memory Architecture. >>>> >>>> Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror >>>> option") >>>> also depends on zone overlap to support sytems with multiple >>>> mirrored ranges. >>>> >>>> Commit c6f03e2903c9 ("mm, memory_hotplug: remove zone restrictions") >>>> embraced overlapped zones for memory hotplug. >>>> >>>> This commit set follows their lead to allow the ZONE_MOVABLE >>>> zone to overlap other zones. Designated Movable Blocks are made >>>> absent from overlapping zones and present within the >>>> ZONE_MOVABLE zone. >>>> >>>> I initially investigated an implementation using a Designated >>>> Movable migrate type in line with comments[1] made by Mel Gorman >>>> regarding a "sticky" MIGRATE_MOVABLE type to avoid using >>>> ZONE_MOVABLE. However, this approach was riskier since it was >>>> much more instrusive on the allocation paths. Ultimately, the >>>> progress made by the memory hotplug folks to expand the >>>> ZONE_MOVABLE functionality convinced me to follow this approach. >>>> >>> >>> Mel, David, does the sub-thread discussion with Doug help ensuring that >>> all of the context is gathered before getting into a more detailed patch >>> review on a patch-by-patch basis? >>> >>> Eventually we may need a fairly firm answer as to whether the proposed >>> approach has any chance of landing upstream in order to either commit to >>> in subsequent iterations of this patch set, or find an alternative. >> >> >> As raised, I'd appreciate if less intrusive alternatives could be >> evaluated (e.g., fake NUMA nodes and being ablee to just use mbind(), >> moving such memory to ZONE_MOVABLE after boot via something like daxctl). > > This is not an option with the environment we have to ultimately fit in > which is Android TV utilizing the GKI kernel which does not enable NUMA > and probably never will, and for similar reasons bringing a whole swath > of user-space tools like daxctl may not be practical either, from both a > logistical perspective (simply getting the tools built with bionic, > accepted etc.) as well as system configuration perspective. Adding feature A because people don't want to (! whoever the "people" are) enable feature B? I hope I don't have to tell you what I think about statements like this :) If feature B is a problem, try stripping it down such that it can be enabled. If it's to hard to configure for your use case, maybe we can extend configuration options to avoid tools like daxctl for some special cases. But of course, only if feature B actually solves the problem. One issue I have with DMB is actual use cases / users / requirements. Maybe requirements are defined somewhere cleanly and I missed them. If we have clear requirements, we can talk about possible solutions. If we have a specific solution, it's harder to talk about requirements. > >> >> I'm not convinced that these intrusive changes are worth it at this >> point. Further, some of the assumptions (ZONE_MOVABLE == user space) are >> not really future proof as I raised. > > I find this patch set reasonably small in contrast to a lot of other mm/ > changes, what did you find intrusive specifically? > > AFAICT, there only assumption that is being made is that ZONE_MOVABLE > contains memory that can be moved, but even if it did not in the future, > there should hopefully be enough opportunities, given a large enough DMB > region to service the allocation requests of its users. I will go back > and read your comment to make sure I don't misunderstand it. Let me clarify what ZONE_MOVABLE can and cannot do: * We cannot assume that specific user space allocations are served from it, neither can we really modify behavior. * We cannot assume that user space allocations won't be migrated off that zone to another zone. * We cannot assume that no other (kernel) allocations will end up on it. * We cannot make specific processes preferably consume memory from it. Designing a feature that relies on any of these assumptions is IMHO wrong. If you want an application to consume memory from a specific address range, there are some possible ways I can see: (1) Model the special memory areas using fake NUMA nodes. e.g., mbind() the applications to these nodes. Use ZONE_MOVABLE to make sure we don't get unmovable allocations. The buddy will take care of it. (2) Use some driver that manages that memory and provides that memory to an application by mmap()'ing it. The buddy won't manage it (no swap, migration ...). DEVDAX is one possible such driver. (3) Use hugetlb and reserve them from the selected memory ranges. Make the application consume these hugetlb pages. For a single node, without a special driver, it gets more complicated: We'd need new way to tell the buddy that these memory ranges are "special". I don't want to use the word "new zone" but that's most likely what it would have to be. Further, one would need a way to specify that only specific allocations should end up on these ranges. Maybe I'm overthinking this. Having clear requirements such that we can try discussing solutions and exploring alternatives would be great.
On 1/4/2023 7:43 AM, Mel Gorman wrote: > On Wed, Dec 14, 2022 at 04:17:35PM -0800, Doug Berger wrote: >> On 11/18/2022 9:05 AM, Mel Gorman wrote: >>> On Wed, Nov 02, 2022 at 03:33:53PM -0700, Doug Berger wrote: [snip] >> I was not familiar with page_alloc.shuffle, but it may very well have a role >> to play here. >> > > It almost certainly does because unlike zones or CMA, it affects how > free lists are arranged. IIRC, the original purpose was about improving > performance of high-speed direct-mapped cache but it also serves a > purpose in this case -- randomising allocations between two channels. > It's still not perfect interleaving but better than none. Agreed. >>>>> A >>>>> major limitation of ZONE_MOVABLE is that there is no way of controlling >>>>> access from userspace to restrict the high-speed memory to a designated >>>>> application, only to all applications in general. The primary interface >>>>> to control access to memory with different characteristics is mempolicies >>>>> which is NUMA orientated, not zone orientated. So, if there is a special >>>>> application that requires exclusive access, it's very difficult to configure >>>>> based on zones. Furthermore, page table pages mapping data located in the >>>>> high-speed region are stored in the slower memory which potentially impacts >>>>> the performance if the working set of the application exceeds TLB reach. >>>>> Finally, while there is mention that Broadcom may have some special >>>>> interface to determine what applications can use the high-speed region, >>>>> it's hardware-specific as opposed to something that belongs in the core mm. >>>>> >>>>> I agree that keeping the high-speed memory in a local node and using "sticky" >>>>> pageblocks or CMA has limitations of its own but in itself, that does not >>>>> justify using ZONE_MOVABLE in my opinion. The statement that ARM can have >>>>> multiple controllers with equal distance and bandwidth (if I'm reading it >>>>> correctly) but places them in different zones.... that's just a bit weird if >>>>> there are no other addressing limitations. It's not obvious why ARM would do >>>>> that, but it also does not matter because it shouldn't be a core mm concern. >>>> >>>> There appears to be some confusion regarding my explanation of multiple >>>> memory controllers on a device like the BCM7278. There is no inherent >>>> performance difference between the two memory controllers and their attached >>>> DRAM. They merely provide the opportunity to perform memory accesses in >>>> parallel for different physical address ranges. The physical address ranges >>>> were selected by the SoC designers for reasons only known to them, but I'm >>>> sure they had no consideration of zones in their decision making. The >>>> selection of zones remains an artifact of the design of Linux. >>>> >>> >>> Ok, so the channels are equal but the channels are not interleaved in >>> hardware so basically you are trying to implement software-based memory >>> channel interleaving? >> >> I suppose that could be a fair characterization of the objective, though the >> approach taken here is very much a "poor man's" approach that attempts to >> improve things without requiring the "heavy lifting" required for a more >> complete solution. >> > > It's still unfortunate that the concept of zones being primarily about > addressing or capability limitations changes. Perhaps arguably, the ZONE_MOVABLE zone continues to be about a capability limitation (i.e. the page allocator cannot use the zone to satisfy requests for non-movable/pinnable memory). This capability limitation has value in different use cases. The hugetlbfs benefits by being able to move data to better compact free memory into higher order free pages. The memory hotplug users benefit by being able to move data before removing memory from the system. A "reusable" reserved memory implementation could benefit from it by being able to move data out of the range when it is reclaimed by the software that owns the reservation. The capability limitation has the follow-on attribute that the zone is prioritized for user-space allocations because the virtual address abstraction of user-space creates the perfect opportunity for physical address independence allowing for movement of data. This is the attribute that is of interest to the multi-channel memory without hardware interleaving use case discussed here rather than the actual movability of the data. The Designated Movable Blocks proposal is a generic mechanism for adding flexibility to determining what memory should be included in the ZONE_MOVABLE zone, and as a result it could support any of these use cases. The memory hotplug developers proposed a similar mechanism early on in their development of what ultimately became the movable_node implementation. > It's also difficult to use as > any user of it has to be very aware of the memory channel configuration of > the machine and know how to match addresses to channels. Information from > zoneinfo on start_pfns, spanned ranges and the like become less useful. It's > relatively minor but splitting the zones also means there is a performance > hit during compaction because pageblock_pfn_to_page is more expensive. I agree that it requires special knowledge of the system to configure for the multi-channel memory without hardware interleaving use case, but that becomes a task for the system administrator that wants to take advantage of the performance benefit of this specific use case. The users don't actually need to be aware of it in this case, and there are no cases where such configuration would occur automatically on systems that were not explicitly interested in it. The memory hotplug developers were able to avoid this complexity using ACPI and SRAT tables, which is why they withdrew their early proposed command line arguments, but those features are not currently available to Broadcom customers. [snip] >> What is of interest to Broadcom customers is to better distribute user space >> accesses across each memory controller to improve the bandwidth available to >> user space dominated work flows. With no ZONE_MOVABLE, the BCM7278 SoC with >> 1GB of memory on each memory controller will place the 1GB on the low >> address memory controller in ZONE_DMA and the 1GB on the high address memory >> controller in ZONE_NORMAL. With this layout movable allocation requests will >> only fallback to the ZONE_DMA (low memory controller) once the ZONE_NORMAL >> (high memory controller) is sufficiently depleted of free memory. >> >> Adding ZONE_MOVABLE memory above ZONE_NORMAL with the current movablecore >> behavior does not improve this situation other than forcing more kernel >> allocations off of the high memory controller. User space allocations are >> even more likely to be on the high memory controller. >> > > But it's a weak promise that interleaving will happen. If only a portion > of ZONE_MOVABLE is used, it might still be all on the same channel. This > might improve over time if enough memory was used and the system was up > for long enough. A "lightly" loaded system is unlikely to see much, if any, benefit from this configuration, but such a system has much less competition for resources. As noted previously, it is the more "heavily" loaded system with multiple parallel user space intensive processes that can benefit by reducing the memory bottleneck created by the biasing of user space allocations to higher addressed zones. The page_alloc.shuffle feature does appear to remove the need for time to pass. > >> The Designated Movable Block mechanism allows ZONE_MOVABLE memory to be >> located on the low memory controller to make it easier for user space >> allocations to land on the low memory controller. If ZONE_MOVABLE is only >> placed on the low memory controller then user space allocations can land in >> ZONE_NORMAL on the high memory controller, but only through fallback after >> ZONE_MOVABLE is sufficiently depleted of free memory which is just the >> reverse of the existing situation. The Designated Movable Block mechanism >> allows ZONE_MOVABLE memory to be located on each memory controller so that >> user space allocations have equal access to each memory controller until the >> ZONE_MOVABLE memory is depleted and fallback to other zones occurs. >> >> To my knowledge Broadcom customers that are currently using the Designated >> Movable Block mechanism are relying on the somewhat random starting and >> stopping of parallel user space processes to produce a more random >> distribution of ZONE_MOVABLE allocations across multiple memory controllers, >> but the page_alloc.shuffle mechanism seems like it would be a good addition >> to promote this randomness. Even better, it appears that page_alloc.shuffle >> is already enabled in the GKI configuration. >> > > The "random starting and stopping of parallel user space processes" is > required for the mechanism to work. It's arbitrary and unknown if the > interleaving happens where as shuffle has an immediate, if random, impact. Yes, page_alloc.shuffle does improve things. > >> You are of course correct that the access patterns make all of the >> difference and it is almost certain that one memory controller or the other >> will be saturated at any given time, but the intent is to increase the >> opportunity to use more of the total bandwidth made available by the >> multiple memory controllers. >> > > And shuffle should also provide that opportunity except it's trivial > to configure and only requires the user to know the memory channels are > not interleaved. The problem with page_alloc.shuffle on its own is that the shuffling can only occur within a zone. As noted for the BCM7278 SoC described above, the low memory controller contains only ZONE_DMA memory and the high memory controller contains only ZONE_NORMAL memory. Shuffling the pages within a zone will not improve the random placement of allocations across the multiple memory controllers unless a zone spans all memory controllers. The creation of Designated Movable Blocks allows a ZONE_MOVABLE zone to be created that spans all memory controllers in the system with an equivalent footprint on each. [snip] >> I experimented with a >> Broadcom BCM7278 system with 1GB on each memory controller (i.e. 2GB total >> memory). The buffers were made large to render data caching meaningless and >> to require several pages to be allocated to populate the buffer. >> >> With V3 of this patch set applied to a 6.1-rc1 kernel I observed these >> results: >> With no movablecore kernel parameter specified: >> # time /tmp/thread_test >> Thread 1 returns: 0 >> Thread 2 returns: 0 >> Thread 3 returns: 0 >> Thread 4 returns: 0 >> >> real 0m4.047s >> user 0m14.183s >> sys 0m1.215s >> >> With this additional kernel parameter "movablecore=600M": >> # time /tmp/thread_test >> Thread 0 returns: 0 >> Thread 1 returns: 0 >> Thread 2 returns: 0 >> Thread 3 returns: 0 >> >> real 0m4.068s >> user 0m14.402s >> sys 0m1.117s >> >> With this additional kernel parameter "movablecore=600M@0x50000000": >> # time /tmp/thread_test >> Thread 0 returns: 0 >> Thread 1 returns: 0 >> Thread 2 returns: 0 >> Thread 3 returns: 0 >> >> real 0m4.010s >> user 0m13.979s >> sys 0m1.070s >> >> However, with these additional kernel parameters >> "movablecore=300M@0x60000000,300M@0x320000000 page_alloc.shuffle=1": >> # time /tmp/thread_test >> Thread 0 returns: 0 >> Thread 1 returns: 0 >> Thread 2 returns: 0 >> Thread 3 returns: 0 >> >> real 0m3.173s >> user 0m11.175s >> sys 0m1.067s >> > > What were the results with just > "movablecore=300M@0x60000000,300M@0x320000000" on its own and > page_alloc.shuffle=1 on its own? > > For shuffle on its own, my expectations are that the results will be > variable, sometimes good and sometimes bad, because it's at the mercy of > the randomisation. It might be slightly improved if the initial top-level > lists were ordered "1, n, 2, n-1, 3, n-2" optionally in __shuffle_zone or > if shuffle_pick_tail was aware of the memory channels but a lot more work > to implement. With the kernel parameters "movablecore=300M@0x60000000,300M@0x320000000" # time /tmp/thread_test Thread 0 returns: 0 Thread 1 returns: 0 Thread 2 returns: 0 Thread 3 returns: 0 real 0m3.562s user 0m12.386s sys 0m1.176s The "movablecore=300M@0x60000000,300M@0x320000000" result is worse than when combined with the shuffle parameter, but may improve over time due to "random starting and stopping of parallel user space processes". With the kernel parameters "page_alloc.shuffle=1" # time /tmp/thread_test Thread 0 returns: 0 Thread 1 returns: 0 Thread 2 returns: 0 Thread 3 returns: 0 real 0m4.056s user 0m14.680s sys 0m1.060s The shuffle on its own result is no better than no movablecore parameter because all of ZONE_NORMAL is on the high memory controller so the pages don't get shuffled between controllers. Happy New Year! Doug
On 1/5/23 05:29, David Hildenbrand wrote: > On 04.01.23 20:00, Florian Fainelli wrote: >> On 1/4/23 07:56, David Hildenbrand wrote: >>> On 04.01.23 00:43, Florian Fainelli wrote: >>>> On 10/20/22 14:53, Doug Berger wrote: >>>>> MOTIVATION: >>>>> Some Broadcom devices (e.g. 7445, 7278) contain multiple memory >>>>> controllers with each mapped in a different address range within >>>>> a Uniform Memory Architecture. Some users of these systems have >>>>> expressed the desire to locate ZONE_MOVABLE memory on each >>>>> memory controller to allow user space intensive processing to >>>>> make better use of the additional memory bandwidth. >>>>> Unfortunately, the historical monotonic layout of zones would >>>>> mean that if the lowest addressed memory controller contains >>>>> ZONE_MOVABLE memory then all of the memory available from >>>>> memory controllers at higher addresses must also be in the >>>>> ZONE_MOVABLE zone. This would force all kernel memory accesses >>>>> onto the lowest addressed memory controller and significantly >>>>> reduce the amount of memory available for non-movable >>>>> allocations. >>>>> >>>>> The main objective of this patch set is therefore to allow a >>>>> block of memory to be designated as part of the ZONE_MOVABLE >>>>> zone where it will always only be used by the kernel page >>>>> allocator to satisfy requests for movable pages. The term >>>>> Designated Movable Block is introduced here to represent such a >>>>> block. The favored implementation allows extension of the >>>>> 'movablecore' kernel parameter to allow specification of a base >>>>> address and support for multiple blocks. The existing >>>>> 'movablecore' mechanisms are retained. >>>>> >>>>> BACKGROUND: >>>>> NUMA architectures support distributing movablecore memory >>>>> across each node, but it is undesirable to introduce the >>>>> overhead and complexities of NUMA on systems that don't have a >>>>> Non-Uniform Memory Architecture. >>>>> >>>>> Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror >>>>> option") >>>>> also depends on zone overlap to support sytems with multiple >>>>> mirrored ranges. >>>>> >>>>> Commit c6f03e2903c9 ("mm, memory_hotplug: remove zone restrictions") >>>>> embraced overlapped zones for memory hotplug. >>>>> >>>>> This commit set follows their lead to allow the ZONE_MOVABLE >>>>> zone to overlap other zones. Designated Movable Blocks are made >>>>> absent from overlapping zones and present within the >>>>> ZONE_MOVABLE zone. >>>>> >>>>> I initially investigated an implementation using a Designated >>>>> Movable migrate type in line with comments[1] made by Mel Gorman >>>>> regarding a "sticky" MIGRATE_MOVABLE type to avoid using >>>>> ZONE_MOVABLE. However, this approach was riskier since it was >>>>> much more instrusive on the allocation paths. Ultimately, the >>>>> progress made by the memory hotplug folks to expand the >>>>> ZONE_MOVABLE functionality convinced me to follow this approach. >>>>> >>>> >>>> Mel, David, does the sub-thread discussion with Doug help ensuring that >>>> all of the context is gathered before getting into a more detailed >>>> patch >>>> review on a patch-by-patch basis? >>>> >>>> Eventually we may need a fairly firm answer as to whether the proposed >>>> approach has any chance of landing upstream in order to either >>>> commit to >>>> in subsequent iterations of this patch set, or find an alternative. >>> >>> >>> As raised, I'd appreciate if less intrusive alternatives could be >>> evaluated (e.g., fake NUMA nodes and being ablee to just use mbind(), >>> moving such memory to ZONE_MOVABLE after boot via something like >>> daxctl). >> >> This is not an option with the environment we have to ultimately fit in >> which is Android TV utilizing the GKI kernel which does not enable NUMA >> and probably never will, and for similar reasons bringing a whole swath >> of user-space tools like daxctl may not be practical either, from both a >> logistical perspective (simply getting the tools built with bionic, >> accepted etc.) as well as system configuration perspective. (looks like I never replied to this email, whoops) > > Adding feature A because people don't want to (! whoever the "people" > are) enable feature B? I hope I don't have to tell you what I think > about statements like this :) It is not just that NUMA is not wanted, it is also not a great fit, the ARM CPU cluster and most peripherals that Linux cares about do have an uniform memory access to the available DRAM controllers/DRAM chips. Only a subset of the peripherals, especially the real-time and high bandwidth ones like video decoders and display that may not be uniformly accessing DRAM. This stems from the fact that the memory controller(s) on the System-on-Chip we work with have a star topology and they schedule the accesses of each DRAM client (CPU, GPU, video decoder, display, Ethernet, PCIe, etc) differently in order to guarantee a certain quality of service. On a system with multiple DRAM controller / DRAM chips, you will typically see video decoder + display instances #0 be serviced by DRAM controller 0, and video decoder + display instance #1 be servied by DRAM controller 1, and this is the only way to allow dual decode + display as they are very bandwidth hungry. The splitting or load balancing is done on a PFN basis, DRAM pages below a certain address are serviced by DRAM controller #0 and those above another cut off are servied by DRAM controller #1. > > If feature B is a problem, try stripping it down such that it can be > enabled. If it's to hard to configure for your use case, maybe we can > extend configuration options to avoid tools like daxctl for some special > cases. I do not see the splitting of the notion of a 'memory node' object away from CONFIG_NUMA going anywhere, and sorry to put that way, but this would be requiring many months for a result that is not even clear, but would be undone anytime someone is not aware of that larger effort. > > But of course, only if feature B actually solves the problem. > > One issue I have with DMB is actual use cases / users / requirements. > Maybe requirements are defined somewhere cleanly and I missed them. That part is entirely fair, the requirements would be as follows: - we need to be able to control precisely across the available DRAM range which specific PFNs fall within specific zones, and the consequence is that we should also be able to have a non-monotonically increasing definition of zones such that there is an appropriate balance between zones and the underlying PFNs / backing DRAM controller instance - device driver(s) should be able to be allocate (via alloc_contig_range() and friends) memory from specific regions of DRAM which should be covered by an underlying zone/fallback/migrate set of heuristics which maximizes the re-use of such memory when the driver is not using it - the underlying zone/fallback/migrate type heuristics should not "excessively" memory in reserve (CMA I am looking at you) but rather should allow for all of the memory in ideal conditions to be "claimed" by the device driver(s) if they desire so - it is acceptable to spend time compacting/reclaiming memory under tight memory pressure since the transitions requiring said driver(s) to allocate are slow path/control events We have other "soft" requirements which are mainly logistical such that: - the least amount of files are changed - there is no need for custom user-space to be running in order to set-up the regions, aka plug & play is highly desirable - there is no dependency upon CONFIG_NUMA in order to simplify the deployment - there is no overhead to the other users of the patch set and the behavior is entirely opt-in > > If we have clear requirements, we can talk about possible solutions. If > we have a specific solution, it's harder to talk about requirements. > >> >> >>> I'm not convinced that these intrusive changes are worth it at this >>> point. Further, some of the assumptions (ZONE_MOVABLE == user space) are >>> not really future proof as I raised. >> >> I find this patch set reasonably small in contrast to a lot of other mm/ >> changes, what did you find intrusive specifically? >> >> AFAICT, there only assumption that is being made is that ZONE_MOVABLE >> contains memory that can be moved, but even if it did not in the future, >> there should hopefully be enough opportunities, given a large enough DMB >> region to service the allocation requests of its users. I will go back >> and read your comment to make sure I don't misunderstand it. > > Let me clarify what ZONE_MOVABLE can and cannot do: > > * We cannot assume that specific user space allocations are served from > it, neither can we really modify behavior. > * We cannot assume that user space allocations won't be migrated off > that zone to another zone. > * We cannot assume that no other (kernel) allocations will end up on it. > * We cannot make specific processes preferably consume memory from it. > > Designing a feature that relies on any of these assumptions is IMHO wrong. > > If you want an application to consume memory from a specific address > range, there are some possible ways I can see: > > (1) Model the special memory areas using fake NUMA nodes. e.g., mbind() > the applications to these nodes. Use ZONE_MOVABLE to make sure we > don't get unmovable allocations. The buddy will take care of it. > (2) Use some driver that manages that memory and provides that memory > to an application by mmap()'ing it. The buddy won't manage it (no > swap, migration ...). DEVDAX is one possible such driver. > (3) Use hugetlb and reserve them from the selected memory ranges. > Make the application consume these hugetlb pages. > > For a single node, without a special driver, it gets more complicated: > We'd need new way to tell the buddy that these memory ranges are > "special". I don't want to use the word "new zone" but that's most > likely what it would have to be. Further, one would need a way to > specify that only specific allocations should end up on these ranges. > > Maybe I'm overthinking this. Having clear requirements such that we can > try discussing solutions and exploring alternatives would be great. This was helpful. We do not need or want to control precisely or exactly where and how applications are allocating memory from, we just need the kernel to do a good enough job at re-using the memory defined in ZONE_MOVABLE when there is no other consumer of the memory residing there. We are perfectly fine with a rather opportunistic and not deterministic approach as it does not require any specific support on the user-space side. For instance using page_alloc.shuffle=1 as Doug showed would be entirely acceptable even if it is fairly naive currently. Hope this helps.
MOTIVATION: Some Broadcom devices (e.g. 7445, 7278) contain multiple memory controllers with each mapped in a different address range within a Uniform Memory Architecture. Some users of these systems have expressed the desire to locate ZONE_MOVABLE memory on each memory controller to allow user space intensive processing to make better use of the additional memory bandwidth. Unfortunately, the historical monotonic layout of zones would mean that if the lowest addressed memory controller contains ZONE_MOVABLE memory then all of the memory available from memory controllers at higher addresses must also be in the ZONE_MOVABLE zone. This would force all kernel memory accesses onto the lowest addressed memory controller and significantly reduce the amount of memory available for non-movable allocations. The main objective of this patch set is therefore to allow a block of memory to be designated as part of the ZONE_MOVABLE zone where it will always only be used by the kernel page allocator to satisfy requests for movable pages. The term Designated Movable Block is introduced here to represent such a block. The favored implementation allows extension of the 'movablecore' kernel parameter to allow specification of a base address and support for multiple blocks. The existing 'movablecore' mechanisms are retained. BACKGROUND: NUMA architectures support distributing movablecore memory across each node, but it is undesirable to introduce the overhead and complexities of NUMA on systems that don't have a Non-Uniform Memory Architecture. Commit 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror option") also depends on zone overlap to support sytems with multiple mirrored ranges. Commit c6f03e2903c9 ("mm, memory_hotplug: remove zone restrictions") embraced overlapped zones for memory hotplug. This commit set follows their lead to allow the ZONE_MOVABLE zone to overlap other zones. Designated Movable Blocks are made absent from overlapping zones and present within the ZONE_MOVABLE zone. I initially investigated an implementation using a Designated Movable migrate type in line with comments[1] made by Mel Gorman regarding a "sticky" MIGRATE_MOVABLE type to avoid using ZONE_MOVABLE. However, this approach was riskier since it was much more instrusive on the allocation paths. Ultimately, the progress made by the memory hotplug folks to expand the ZONE_MOVABLE functionality convinced me to follow this approach. Changes in v3: - removed OTHER OPPORTUNITIES and NOTES from this cover letter. - prevent the creation of empty zones instead of adding extra info to zoneinfo. - size the ZONE_MOVABLE span to the minimum necessary to cover pages within the zone to be more intuitive. - removed "real" from variable names that were consolidated. - rebased to akpm-mm/master (i.e. Linux 6.1-rc1). Changes in v2: - first three commits upstreamed separately [3], [4], and [5]. - commits 04-06 submitted separately [6]. - Corrected errors "Reported-by: kernel test robot <lkp@intel.com>" - Deferred commits after 15 to simplify review of the base functionality. - minor reorganization of commit 13. v2: https://lore.kernel.org/linux-mm/20220928223301.375229-1-opendmb@gmail.com/ v1: https://lore.kernel.org/linux-mm/20220913195508.3511038-1-opendmb@gmail.com/ [1] https://lore.kernel.org/lkml/20160428103927.GM2858@techsingularity.net/ [2] https://lore.kernel.org/lkml/1401260672-28339-1-git-send-email-iamjoonsoo.kim@lge.com [3] https://lore.kernel.org/linux-mm/20220914023913.1855924-1-zi.yan@sent.com [4] https://lore.kernel.org/linux-mm/20220823030209.57434-2-linmiaohe@huawei.com [5] https://lore.kernel.org/linux-mm/20220914190917.3517663-1-opendmb@gmail.com [6] https://lore.kernel.org/linux-mm/20220921223639.1152392-1-opendmb@gmail.com/ Doug Berger (9): lib/show_mem.c: display MovableOnly mm/page_alloc: calculate node_spanned_pages from pfns mm/page_alloc: prevent creation of empty zones mm/page_alloc.c: allow oversized movablecore mm/page_alloc: introduce init_reserved_pageblock() memblock: introduce MEMBLOCK_MOVABLE flag mm/dmb: Introduce Designated Movable Blocks mm/page_alloc: make alloc_contig_pages DMB aware mm/page_alloc: allow base for movablecore .../admin-guide/kernel-parameters.txt | 14 +- include/linux/dmb.h | 29 +++ include/linux/gfp.h | 5 +- include/linux/memblock.h | 8 + lib/show_mem.c | 2 +- mm/Kconfig | 12 ++ mm/Makefile | 1 + mm/cma.c | 15 +- mm/dmb.c | 91 +++++++++ mm/memblock.c | 30 ++- mm/page_alloc.c | 188 +++++++++++++----- 11 files changed, 338 insertions(+), 57 deletions(-) create mode 100644 include/linux/dmb.h create mode 100644 mm/dmb.c