| Message ID | 20240530201616.1316526-1-almasrymina@google.com |
|---|---|
| Headers | show |
| Series | Device Memory TCP | expand |
On 2024-05-30 13:16, Mina Almasry wrote: [...] > +err_start_queue: > + /* Restarting the queue with old_mem should be successful as we haven't > + * changed any of the queue configuration, and there is not much we can > + * do to recover from a failure here. > + * > + * WARN if the we fail to recover the old rx queue, and at least free > + * old_mem so we don't also leak that. > + */ > + if (dev->queue_mgmt_ops->ndo_queue_start(dev, old_mem, rxq_idx)) { > + WARN(1, > + "Failed to restart old queue in error path. RX queue %d may be unhealthy.", > + rxq_idx); > + dev->queue_mgmt_ops->ndo_queue_mem_free(dev, &old_mem); This should be ->ndo_queue_mem_free(dev, old_mem).
On Thu, May 30, 2024 at 08:16:12PM +0000, Mina Almasry wrote: > Add documentation outlining the usage and details of devmem TCP. > > Signed-off-by: Mina Almasry <almasrymina@google.com> > The doc LGTM, thanks! Reviewed-by: Bagas Sanjaya <bagasdotme@gmail.com>
On 5/30/24 21:16, Mina Almasry wrote: > Add netdev_rx_queue_restart() function to netdev_rx_queue.h > > Signed-off-by: David Wei <dw@davidwei.uk> > Signed-off-by: Mina Almasry <almasrymina@google.com> > > --- ... > diff --git a/net/core/netdev_rx_queue.c b/net/core/netdev_rx_queue.c > new file mode 100644 > index 0000000000000..b3899358e5a9c > --- /dev/null > +++ b/net/core/netdev_rx_queue.c > @@ -0,0 +1,74 @@ > +// SPDX-License-Identifier: GPL-2.0-or-later > + > +#include <linux/netdevice.h> > +#include <net/netdev_queues.h> > +#include <net/netdev_rx_queue.h> > + > +int netdev_rx_queue_restart(struct net_device *dev, unsigned int rxq_idx) > +{ > + void *new_mem, *old_mem; > + int err; I believe it should also do: if (!dev->queue_mgmt_ops) return -EOPNOTSUPP; > + > + if (!dev->queue_mgmt_ops->ndo_queue_stop || > + !dev->queue_mgmt_ops->ndo_queue_mem_free || > + !dev->queue_mgmt_ops->ndo_queue_mem_alloc || > + !dev->queue_mgmt_ops->ndo_queue_start) > + return -EOPNOTSUPP; > + > + DEBUG_NET_WARN_ON_ONCE(!rtnl_is_locked());
On Thu, 2024-05-30 at 20:16 +0000, Mina Almasry wrote: > diff --git a/net/core/devmem.c b/net/core/devmem.c > index d82f92d7cf9ce..d5fac8edf621d 100644 > --- a/net/core/devmem.c > +++ b/net/core/devmem.c > @@ -32,6 +32,14 @@ static void net_devmem_dmabuf_free_chunk_owner(struct gen_pool *genpool, > kfree(owner); > } > > +static inline dma_addr_t net_devmem_get_dma_addr(const struct net_iov *niov) Minor nit: please no 'inline' keyword in c files. Thanks, Paolo
On Tue, 04 Jun 2024 12:13:15 +0200 Paolo Abeni <pabeni@redhat.com> wrote: > On Thu, 2024-05-30 at 20:16 +0000, Mina Almasry wrote: > > diff --git a/net/core/devmem.c b/net/core/devmem.c > > index d82f92d7cf9ce..d5fac8edf621d 100644 > > --- a/net/core/devmem.c > > +++ b/net/core/devmem.c > > @@ -32,6 +32,14 @@ static void net_devmem_dmabuf_free_chunk_owner(struct gen_pool *genpool, > > kfree(owner); > > } > > > > +static inline dma_addr_t net_devmem_get_dma_addr(const struct net_iov *niov) > > Minor nit: please no 'inline' keyword in c files. I'm curious. Is this a networking rule? I use 'inline' in my C code all the time. -- Steve
On Tue, Jun 04, 2024 at 12:15:51PM -0400, Steven Rostedt wrote: > On Tue, 04 Jun 2024 12:13:15 +0200 > Paolo Abeni <pabeni@redhat.com> wrote: > > > On Thu, 2024-05-30 at 20:16 +0000, Mina Almasry wrote: > > > diff --git a/net/core/devmem.c b/net/core/devmem.c > > > index d82f92d7cf9ce..d5fac8edf621d 100644 > > > --- a/net/core/devmem.c > > > +++ b/net/core/devmem.c > > > @@ -32,6 +32,14 @@ static void net_devmem_dmabuf_free_chunk_owner(struct gen_pool *genpool, > > > kfree(owner); > > > } > > > > > > +static inline dma_addr_t net_devmem_get_dma_addr(const struct net_iov *niov) > > > > Minor nit: please no 'inline' keyword in c files. > > I'm curious. Is this a networking rule? I use 'inline' in my C code all the > time. It mostly comes from Documentation/process/coding-style.rst: 15) The inline disease ---------------------- There appears to be a common misperception that gcc has a magic "make me faster" speedup option called ``inline``. While the use of inlines can be appropriate (for example as a means of replacing macros, see Chapter 12), it very often is not. Abundant use of the inline keyword leads to a much bigger kernel, which in turn slows the system as a whole down, due to a bigger icache footprint for the CPU and simply because there is less memory available for the pagecache. Just think about it; a pagecache miss causes a disk seek, which easily takes 5 milliseconds. There are a LOT of cpu cycles that can go into these 5 milliseconds. A reasonable rule of thumb is to not put inline at functions that have more than 3 lines of code in them. An exception to this rule are the cases where a parameter is known to be a compiletime constant, and as a result of this constantness you *know* the compiler will be able to optimize most of your function away at compile time. For a good example of this later case, see the kmalloc() inline function. Often people argue that adding inline to functions that are static and used only once is always a win since there is no space tradeoff. While this is technically correct, gcc is capable of inlining these automatically without help, and the maintenance issue of removing the inline when a second user appears outweighs the potential value of the hint that tells gcc to do something it would have done anyway. Jason
On Tue, 4 Jun 2024 13:31:58 -0300 Jason Gunthorpe <jgg@ziepe.ca> wrote: > On Tue, Jun 04, 2024 at 12:15:51PM -0400, Steven Rostedt wrote: > > On Tue, 04 Jun 2024 12:13:15 +0200 > > Paolo Abeni <pabeni@redhat.com> wrote: > > > > > On Thu, 2024-05-30 at 20:16 +0000, Mina Almasry wrote: > > > > diff --git a/net/core/devmem.c b/net/core/devmem.c > > > > index d82f92d7cf9ce..d5fac8edf621d 100644 > > > > --- a/net/core/devmem.c > > > > +++ b/net/core/devmem.c > > > > @@ -32,6 +32,14 @@ static void net_devmem_dmabuf_free_chunk_owner(struct gen_pool *genpool, > > > > kfree(owner); > > > > } > > > > > > > > +static inline dma_addr_t net_devmem_get_dma_addr(const struct net_iov *niov) > > > > > > Minor nit: please no 'inline' keyword in c files. > > > > I'm curious. Is this a networking rule? I use 'inline' in my C code all the > > time. > > It mostly comes from Documentation/process/coding-style.rst: > > 15) The inline disease > ---------------------- > > There appears to be a common misperception that gcc has a magic "make me > faster" speedup option called ``inline``. While the use of inlines can be > appropriate (for example as a means of replacing macros, see Chapter 12), it > very often is not. Abundant use of the inline keyword leads to a much bigger > kernel, which in turn slows the system as a whole down, due to a bigger > icache footprint for the CPU and simply because there is less memory > available for the pagecache. Just think about it; a pagecache miss causes a > disk seek, which easily takes 5 milliseconds. There are a LOT of cpu cycles > that can go into these 5 milliseconds. > > A reasonable rule of thumb is to not put inline at functions that have more > than 3 lines of code in them. An exception to this rule are the cases where > a parameter is known to be a compiletime constant, and as a result of this > constantness you *know* the compiler will be able to optimize most of your > function away at compile time. For a good example of this later case, see > the kmalloc() inline function. > > Often people argue that adding inline to functions that are static and used > only once is always a win since there is no space tradeoff. While this is > technically correct, gcc is capable of inlining these automatically without > help, and the maintenance issue of removing the inline when a second user > appears outweighs the potential value of the hint that tells gcc to do > something it would have done anyway. > Interesting, as I sped up the ftrace ring buffer by a substantial amount by adding strategic __always_inline, noinline, likely() and unlikely() throughout the code. It had to do with what was considered the fast path and slow path, and not actually the size of the function. gcc got it horribly wrong. -- Steve
> Interesting, as I sped up the ftrace ring buffer by a substantial amount by > adding strategic __always_inline, noinline, likely() and unlikely() > throughout the code. It had to do with what was considered the fast path > and slow path, and not actually the size of the function. gcc got it > horribly wrong. And what did the compiler people say when you reported gcc was getting it wrong? Our assumption is, the compiler is better than a human at deciding this. Or at least, a human who does not spend a long time profiling and tuning. If this assumption is not true, we probably should be trying to figure out why, and improving the compiler when possible. That will benefit everybody. Andrew
On Wed, 5 Jun 2024 01:44:37 +0200 Andrew Lunn <andrew@lunn.ch> wrote: > > Interesting, as I sped up the ftrace ring buffer by a substantial amount by > > adding strategic __always_inline, noinline, likely() and unlikely() > > throughout the code. It had to do with what was considered the fast path > > and slow path, and not actually the size of the function. gcc got it > > horribly wrong. > > And what did the compiler people say when you reported gcc was getting > it wrong? > > Our assumption is, the compiler is better than a human at deciding > this. Or at least, a human who does not spend a long time profiling > and tuning. If this assumption is not true, we probably should be > trying to figure out why, and improving the compiler when > possible. That will benefit everybody. > How is the compiler going to know which path is going to be taken the most? There's two main paths in the ring buffer logic. One when an event stays on the sub-buffer, the other when the event crosses over to a new sub buffer. As there's 100s of events that happen on the same sub-buffer for every one time there's a cross over, I optimized the paths that stayed on the sub-buffer, which caused the time for those events to go from 250ns down to 150 ns!. That's a 40% speed up. I added the unlikely/likely and 'always_inline' and 'noinline' paths to make sure the "staying on the buffer" path was always the hot path, and keeping it tight in cache. How is a compiler going to know that? -- Steve
> How is the compiler going to know which path is going to be taken the most? > There's two main paths in the ring buffer logic. One when an event stays on > the sub-buffer, the other when the event crosses over to a new sub buffer. > As there's 100s of events that happen on the same sub-buffer for every one > time there's a cross over, I optimized the paths that stayed on the > sub-buffer, which caused the time for those events to go from 250ns down to > 150 ns!. That's a 40% speed up. > > I added the unlikely/likely and 'always_inline' and 'noinline' paths to > make sure the "staying on the buffer" path was always the hot path, and > keeping it tight in cache. > > How is a compiler going to know that? It might have some heuristics to try to guess unlikely/likely, but that is not what we are talking about here. How much difference did 'always_inline' and 'noinline' make? Hopefully the likely is enough of a clue it should prefer to inline whatever is in that branch, where as for the unlikely case it can do a function call. But compilers is not my thing, which is why i would reach out to the compiler people and ask them, is it expected to get this wrong, could it be made better? Andrew
On Wed, 5 Jun 2024 02:52:29 +0200 Andrew Lunn <andrew@lunn.ch> wrote: > > How is a compiler going to know that? > > It might have some heuristics to try to guess unlikely/likely, but > that is not what we are talking about here. > > How much difference did 'always_inline' and 'noinline' make? Hopefully > the likely is enough of a clue it should prefer to inline whatever is > in that branch, where as for the unlikely case it can do a function > call. Perhaps, but one of the issues was that I have lots of small functions that are used all over the place, and gcc tends to change them to function calls, instead of duplicating them. I did this analysis back in 2016, so maybe it became better. > > But compilers is not my thing, which is why i would reach out to the > compiler people and ask them, is it expected to get this wrong, could > it be made better? Well, I actually do work with the compiler folks, and we are actually trying to get a session at GNU Cauldron where Linux kernel folks can talk with the gcc compiler folks. I've stared at so many objdump outputs, that I can now pretty much see the assembly that my C code makes ;-) -- Steve
On Tue, 2024-06-04 at 20:27 -0400, Steven Rostedt wrote: > On Wed, 5 Jun 2024 01:44:37 +0200 > Andrew Lunn <andrew@lunn.ch> wrote: > > > > Interesting, as I sped up the ftrace ring buffer by a substantial amount by > > > adding strategic __always_inline, noinline, likely() and unlikely() > > > throughout the code. It had to do with what was considered the fast path > > > and slow path, and not actually the size of the function. gcc got it > > > horribly wrong. > > > > And what did the compiler people say when you reported gcc was getting > > it wrong? > > > > Our assumption is, the compiler is better than a human at deciding > > this. Or at least, a human who does not spend a long time profiling > > and tuning. If this assumption is not true, we probably should be > > trying to figure out why, and improving the compiler when > > possible. That will benefit everybody. > > > > How is the compiler going to know which path is going to be taken the most? > There's two main paths in the ring buffer logic. One when an event stays on > the sub-buffer, the other when the event crosses over to a new sub buffer. > As there's 100s of events that happen on the same sub-buffer for every one > time there's a cross over, I optimized the paths that stayed on the > sub-buffer, which caused the time for those events to go from 250ns down to > 150 ns!. That's a 40% speed up. > > I added the unlikely/likely and 'always_inline' and 'noinline' paths to > make sure the "staying on the buffer" path was always the hot path, and > keeping it tight in cache. > > How is a compiler going to know that? > > -- Steve > Isn't this basically a perfect example of something where profile guided optimization should work? Thanks, Niklas
v10: https://patchwork.kernel.org/project/netdevbpf/list/?series=852422&state=* ==== Major Changes: -------------- v9 was sent right before the merge window closed (sorry!). v10 is almost a re-send of the series now that the merge window re-opened. Only rebased to latest net-next and addressed some minor iterative comments received on v9. As usual, the full devmem TCP changes including the full GVE driver implementation is here: https://github.com/mina/linux/commits/tcpdevmem-v10/ Detailed changelog: ------------------- - Fixed tokens leaking in DONTNEED setsockopt (Nikolay). - Moved net_iov_dma_addr() to devmem.c and made it a devmem specific helpers (David). - Rename hook alloc_pages to alloc_netmems as alloc_pages is now preprocessor macro defined and causes a build error. v9: === Major Changes: -------------- GVE queue API has been merged. Submitting this version as non-RFC after rebasing on top of the merged API, and dropped the out of tree queue API I was carrying on github. Addressed the little feedback v8 has received. Detailed changelog: ------------------ - Added new patch from David Wei to this series for netdev_rx_queue_restart() - Fixed sparse error. - Removed CONFIG_ checks in netmem_is_net_iov() - Flipped skb->readable to skb->unreadable - Minor fixes to selftests & docs. RFC v8: ======= Major Changes: -------------- - Fixed build error generated by patch-by-patch build. - Applied docs suggestions from Randy. RFC v7: ======= Major Changes: -------------- This revision largely rebases on top of net-next and addresses the feedback RFCv6 received from folks, namely Jakub, Yunsheng, Arnd, David, & Pavel. The series remains in RFC because the queue-API ndos defined in this series are not yet implemented. I have a GVE implementation I carry out of tree for my testing. A upstreamable GVE implementation is in the works. Aside from that, in my estimation all the patches are ready for review/merge. Please do take a look. As usual the full devmem TCP changes including the full GVE driver implementation is here: https://github.com/mina/linux/commits/tcpdevmem-v7/ Detailed changelog: - Use admin-perm in netlink API. - Addressed feedback from Jakub with regards to netlink API implementation. - Renamed devmem.c functions to something more appropriate for that file. - Improve the performance seen through the page_pool benchmark. - Fix the value definition of all the SO_DEVMEM_* uapi. - Various fixes to documentation. Perf - page-pool benchmark: --------------------------- Improved performance of bench_page_pool_simple.ko tests compared to v6: https://pastebin.com/raw/v5dYRg8L net-next base: 8 cycle fast path. RFC v6: 10 cycle fast path. RFC v7: 9 cycle fast path. RFC v7 with CONFIG_DMA_SHARED_BUFFER disabled: 8 cycle fast path, same as baseline. Perf - Devmem TCP benchmark: --------------------- Perf is about the same regardless of the changes in v7, namely the removal of the static_branch_unlikely to improve the page_pool benchmark performance: 189/200gbps bi-directional throughput with RX devmem TCP and regular TCP TX i.e. ~95% line rate. RFC v6: ======= Major Changes: -------------- This revision largely rebases on top of net-next and addresses the little feedback RFCv5 received. The series remains in RFC because the queue-API ndos defined in this series are not yet implemented. I have a GVE implementation I carry out of tree for my testing. A upstreamable GVE implementation is in the works. Aside from that, in my estimation all the patches are ready for review/merge. Please do take a look. As usual the full devmem TCP changes including the full GVE driver implementation is here: https://github.com/mina/linux/commits/tcpdevmem-v6/ This version also comes with some performance data recorded in the cover letter (see below changelog). Detailed changelog: - Rebased on top of the merged netmem_ref changes. - Converted skb->dmabuf to skb->readable (Pavel). Pavel's original suggestion was to remove the skb->dmabuf flag entirely, but when I looked into it closely, I found the issue that if we remove the flag we have to dereference the shinfo(skb) pointer to obtain the first frag to tell whether an skb is readable or not. This can cause a performance regression if it dirties the cache line when the shinfo(skb) was not really needed. Instead, I converted the skb->dmabuf flag into a generic skb->readable flag which can be re-used by io_uring 0-copy RX. - Squashed a few locking optimizations from Eric Dumazet in the RX path and the DEVMEM_DONTNEED setsockopt. - Expanded the tests a bit. Added validation for invalid scenarios and added some more coverage. Perf - page-pool benchmark: --------------------------- bench_page_pool_simple.ko tests with and without these changes: https://pastebin.com/raw/ncHDwAbn AFAIK the number that really matters in the perf tests is the 'tasklet_page_pool01_fast_path Per elem'. This one measures at about 8 cycles without the changes but there is some 1 cycle noise in some results. With the patches this regresses to 9 cycles with the changes but there is 1 cycle noise occasionally running this test repeatedly. Lastly I tried disable the static_branch_unlikely() in netmem_is_net_iov() check. To my surprise disabling the static_branch_unlikely() check reduces the fast path back to 8 cycles, but the 1 cycle noise remains. Perf - Devmem TCP benchmark: --------------------- 189/200gbps bi-directional throughput with RX devmem TCP and regular TCP TX i.e. ~95% line rate. Major changes in RFC v5: ======================== 1. Rebased on top of 'Abstract page from net stack' series and used the new netmem type to refer to LSB set pointers instead of re-using struct page. 2. Downgraded this series back to RFC and called it RFC v5. This is because this series is now dependent on 'Abstract page from net stack'[1] and the queue API. Both are removed from the series to reduce the patch # and those bits are fairly independent or pre-requisite work. 3. Reworked the page_pool devmem support to use netmem and for some more unified handling. 4. Reworked the reference counting of net_iov (renamed from page_pool_iov) to use pp_ref_count for refcounting. The full changes including the dependent series and GVE page pool support is here: https://github.com/mina/linux/commits/tcpdevmem-rfcv5/ [1] https://patchwork.kernel.org/project/netdevbpf/list/?series=810774 Major changes in v1: ==================== 1. Implemented MVP queue API ndos to remove the userspace-visible driver reset. 2. Fixed issues in the napi_pp_put_page() devmem frag unref path. 3. Removed RFC tag. Many smaller addressed comments across all the patches (patches have individual change log). Full tree including the rest of the GVE driver changes: https://github.com/mina/linux/commits/tcpdevmem-v1 Changes in RFC v3: ================== 1. Pulled in the memory-provider dependency from Jakub's RFC[1] to make the series reviewable and mergeable. 2. Implemented multi-rx-queue binding which was a todo in v2. 3. Fix to cmsg handling. The sticking point in RFC v2[2] was the device reset required to refill the device rx-queues after the dmabuf bind/unbind. The solution suggested as I understand is a subset of the per-queue management ops Jakub suggested or similar: https://lore.kernel.org/netdev/20230815171638.4c057dcd@kernel.org/ This is not addressed in this revision, because: 1. This point was discussed at netconf & netdev and there is openness to using the current approach of requiring a device reset. 2. Implementing individual queue resetting seems to be difficult for my test bed with GVE. My prototype to test this ran into issues with the rx-queues not coming back up properly if reset individually. At the moment I'm unsure if it's a mistake in the POC or a genuine issue in the virtualization stack behind GVE, which currently doesn't test individual rx-queue restart. 3. Our usecases are not bothered by requiring a device reset to refill the buffer queues, and we'd like to support NICs that run into this limitation with resetting individual queues. My thought is that drivers that have trouble with per-queue configs can use the support in this series, while drivers that support new netdev ops to reset individual queues can automatically reset the queue as part of the dma-buf bind/unbind. The same approach with device resets is presented again for consideration with other sticking points addressed. This proposal includes the rx devmem path only proposed for merge. For a snapshot of my entire tree which includes the GVE POC page pool support & device memory support: https://github.com/torvalds/linux/compare/master...mina:linux:tcpdevmem-v3 [1] https://lore.kernel.org/netdev/f8270765-a27b-6ccf-33ea-cda097168d79@redhat.com/T/ [2] https://lore.kernel.org/netdev/CAHS8izOVJGJH5WF68OsRWFKJid1_huzzUK+hpKbLcL4pSOD1Jw@mail.gmail.com/T/ Changes in RFC v2: ================== The sticking point in RFC v1[1] was the dma-buf pages approach we used to deliver the device memory to the TCP stack. RFC v2 is a proof-of-concept that attempts to resolve this by implementing scatterlist support in the networking stack, such that we can import the dma-buf scatterlist directly. This is the approach proposed at a high level here[2]. Detailed changes: 1. Replaced dma-buf pages approach with importing scatterlist into the page pool. 2. Replace the dma-buf pages centric API with a netlink API. 3. Removed the TX path implementation - there is no issue with implementing the TX path with scatterlist approach, but leaving out the TX path makes it easier to review. 4. Functionality is tested with this proposal, but I have not conducted perf testing yet. I'm not sure there are regressions, but I removed perf claims from the cover letter until they can be re-confirmed. 5. Added Signed-off-by: contributors to the implementation. 6. Fixed some bugs with the RX path since RFC v1. Any feedback welcome, but specifically the biggest pending questions needing feedback IMO are: 1. Feedback on the scatterlist-based approach in general. 2. Netlink API (Patch 1 & 2). 3. Approach to handle all the drivers that expect to receive pages from the page pool (Patch 6). [1] https://lore.kernel.org/netdev/dfe4bae7-13a0-3c5d-d671-f61b375cb0b4@gmail.com/T/ [2] https://lore.kernel.org/netdev/CAHS8izPm6XRS54LdCDZVd0C75tA1zHSu6jLVO8nzTLXCc=H7Nw@mail.gmail.com/ ================== * TL;DR: Device memory TCP (devmem TCP) is a proposal for transferring data to and/or from device memory efficiently, without bouncing the data to a host memory buffer. * Problem: A large amount of data transfers have device memory as the source and/or destination. Accelerators drastically increased the volume of such transfers. Some examples include: - ML accelerators transferring large amounts of training data from storage into GPU/TPU memory. In some cases ML training setup time can be as long as 50% of TPU compute time, improving data transfer throughput & efficiency can help improving GPU/TPU utilization. - Distributed training, where ML accelerators, such as GPUs on different hosts, exchange data among them. - Distributed raw block storage applications transfer large amounts of data with remote SSDs, much of this data does not require host processing. Today, the majority of the Device-to-Device data transfers the network are implemented as the following low level operations: Device-to-Host copy, Host-to-Host network transfer, and Host-to-Device copy. The implementation is suboptimal, especially for bulk data transfers, and can put significant strains on system resources, such as host memory bandwidth, PCIe bandwidth, etc. One important reason behind the current state is the kernel’s lack of semantics to express device to network transfers. * Proposal: In this patch series we attempt to optimize this use case by implementing socket APIs that enable the user to: 1. send device memory across the network directly, and 2. receive incoming network packets directly into device memory. Packet _payloads_ go directly from the NIC to device memory for receive and from device memory to NIC for transmit. Packet _headers_ go to/from host memory and are processed by the TCP/IP stack normally. The NIC _must_ support header split to achieve this. Advantages: - Alleviate host memory bandwidth pressure, compared to existing network-transfer + device-copy semantics. - Alleviate PCIe BW pressure, by limiting data transfer to the lowest level of the PCIe tree, compared to traditional path which sends data through the root complex. * Patch overview: ** Part 1: netlink API Gives user ability to bind dma-buf to an RX queue. ** Part 2: scatterlist support Currently the standard for device memory sharing is DMABUF, which doesn't generate struct pages. On the other hand, networking stack (skbs, drivers, and page pool) operate on pages. We have 2 options: 1. Generate struct pages for dmabuf device memory, or, 2. Modify the networking stack to process scatterlist. Approach #1 was attempted in RFC v1. RFC v2 implements approach #2. ** part 3: page pool support We piggy back on page pool memory providers proposal: https://github.com/kuba-moo/linux/tree/pp-providers It allows the page pool to define a memory provider that provides the page allocation and freeing. It helps abstract most of the device memory TCP changes from the driver. ** part 4: support for unreadable skb frags Page pool iovs are not accessible by the host; we implement changes throughput the networking stack to correctly handle skbs with unreadable frags. ** Part 5: recvmsg() APIs We define user APIs for the user to send and receive device memory. Not included with this series is the GVE devmem TCP support, just to simplify the review. Code available here if desired: https://github.com/mina/linux/tree/tcpdevmem This series is built on top of net-next with Jakub's pp-providers changes cherry-picked. * NIC dependencies: 1. (strict) Devmem TCP require the NIC to support header split, i.e. the capability to split incoming packets into a header + payload and to put each into a separate buffer. Devmem TCP works by using device memory for the packet payload, and host memory for the packet headers. 2. (optional) Devmem TCP works better with flow steering support & RSS support, i.e. the NIC's ability to steer flows into certain rx queues. This allows the sysadmin to enable devmem TCP on a subset of the rx queues, and steer devmem TCP traffic onto these queues and non devmem TCP elsewhere. The NIC I have access to with these properties is the GVE with DQO support running in Google Cloud, but any NIC that supports these features would suffice. I may be able to help reviewers bring up devmem TCP on their NICs. * Testing: The series includes a udmabuf kselftest that show a simple use case of devmem TCP and validates the entire data path end to end without a dependency on a specific dmabuf provider. ** Test Setup Kernel: net-next with this series and memory provider API cherry-picked locally. Hardware: Google Cloud A3 VMs. NIC: GVE with header split & RSS & flow steering support. Cc: Pavel Begunkov <asml.silence@gmail.com> Cc: David Wei <dw@davidwei.uk> Cc: Jason Gunthorpe <jgg@ziepe.ca> Cc: Yunsheng Lin <linyunsheng@huawei.com> Cc: Shailend Chand <shailend@google.com> Cc: Harshitha Ramamurthy <hramamurthy@google.com> Cc: Shakeel Butt <shakeel.butt@linux.dev> Cc: Jeroen de Borst <jeroendb@google.com> Cc: Praveen Kaligineedi <pkaligineedi@google.com> Jakub Kicinski (1): net: page_pool: create hooks for custom page providers Mina Almasry (13): netdev: add netdev_rx_queue_restart() net: netdev netlink api to bind dma-buf to a net device netdev: support binding dma-buf to netdevice netdev: netdevice devmem allocator page_pool: convert to use netmem page_pool: devmem support memory-provider: dmabuf devmem memory provider net: support non paged skb frags net: add support for skbs with unreadable frags tcp: RX path for devmem TCP net: add SO_DEVMEM_DONTNEED setsockopt to release RX frags net: add devmem TCP documentation selftests: add ncdevmem, netcat for devmem TCP Documentation/netlink/specs/netdev.yaml | 57 +++ Documentation/networking/devmem.rst | 258 +++++++++++ Documentation/networking/index.rst | 1 + arch/alpha/include/uapi/asm/socket.h | 6 + arch/mips/include/uapi/asm/socket.h | 6 + arch/parisc/include/uapi/asm/socket.h | 6 + arch/sparc/include/uapi/asm/socket.h | 6 + include/linux/skbuff.h | 61 ++- include/linux/skbuff_ref.h | 11 +- include/linux/socket.h | 1 + include/net/devmem.h | 124 ++++++ include/net/netdev_rx_queue.h | 5 + include/net/netmem.h | 208 ++++++++- include/net/page_pool/helpers.h | 153 +++++-- include/net/page_pool/types.h | 33 +- include/net/sock.h | 2 + include/net/tcp.h | 5 +- include/trace/events/page_pool.h | 29 +- include/uapi/asm-generic/socket.h | 6 + include/uapi/linux/netdev.h | 19 + include/uapi/linux/uio.h | 17 + net/bpf/test_run.c | 5 +- net/core/Makefile | 3 +- net/core/datagram.c | 6 + net/core/dev.c | 6 +- net/core/devmem.c | 384 +++++++++++++++++ net/core/gro.c | 8 +- net/core/netdev-genl-gen.c | 23 + net/core/netdev-genl-gen.h | 6 + net/core/netdev-genl.c | 103 +++++ net/core/netdev_rx_queue.c | 74 ++++ net/core/page_pool.c | 368 +++++++++------- net/core/skbuff.c | 83 +++- net/core/sock.c | 61 +++ net/ipv4/esp4.c | 3 +- net/ipv4/tcp.c | 254 ++++++++++- net/ipv4/tcp_input.c | 13 +- net/ipv4/tcp_ipv4.c | 10 + net/ipv4/tcp_minisocks.c | 2 + net/ipv4/tcp_output.c | 5 +- net/ipv6/esp6.c | 3 +- net/packet/af_packet.c | 4 +- tools/include/uapi/linux/netdev.h | 19 + tools/testing/selftests/net/.gitignore | 1 + tools/testing/selftests/net/Makefile | 5 + tools/testing/selftests/net/ncdevmem.c | 542 ++++++++++++++++++++++++ 46 files changed, 2738 insertions(+), 267 deletions(-) create mode 100644 Documentation/networking/devmem.rst create mode 100644 include/net/devmem.h create mode 100644 net/core/devmem.c create mode 100644 net/core/netdev_rx_queue.c create mode 100644 tools/testing/selftests/net/ncdevmem.c