Merge branch 'per-epoll-context-busy-poll'
Joe Damato says: ==================== Per epoll context busy poll support Greetings: Welcome to v8. TL;DR This builds on commit bf3b9f63 ("epoll: Add busy poll support to epoll with socket fds.") by allowing user applications to enable epoll-based busy polling, set a busy poll packet budget, and enable or disable prefer busy poll on a per epoll context basis. This makes epoll-based busy polling much more usable for user applications than the current system-wide sysctl and hardcoded budget. To allow for this, two ioctls have been added for epoll contexts for getting and setting a new struct, struct epoll_params. ioctl was chosen vs a new syscall after reviewing a suggestion by Willem de Bruijn [1]. I am open to using a new syscall instead of an ioctl, but it seemed that: - Busy poll affects all existing epoll_wait and epoll_pwait variants in the same way, so new verions of many syscalls might be needed. It seems much simpler for users to use the correct epoll_wait/epoll_pwait for their app and add a call to ioctl to enable or disable busy poll as needed. This also probably means less work to get an existing epoll app using busy poll. - previously added epoll_pwait2 helped to bring epoll closer to existing syscalls (like pselect and ppoll) and this busy poll change reflected as a new syscall would not have the same effect. Note: patch 1/4 as of v4 uses an or (||) instead of an xor. I thought about it some more and I realized that if the user enables both the per-epoll context setting and the system wide sysctl, then busy poll should be enabled and not disabled. Using xor doesn't seem to make much sense after thinking through this a bit. Longer explanation: Presently epoll has support for a very useful form of busy poll based on the incoming NAPI ID (see also: SO_INCOMING_NAPI_ID [2]). This form of busy poll allows epoll_wait to drive NAPI packet processing which allows for a few interesting user application designs which can reduce latency and also potentially improve L2/L3 cache hit rates by deferring NAPI until userland has finished its work. The documentation available on this is, IMHO, a bit confusing so please allow me to explain how one might use this: 1. Ensure each application thread has its own epoll instance mapping 1-to-1 with NIC RX queues. An n-tuple filter would likely be used to direct connections with specific dest ports to these queues. 2. Optionally: Setup IRQ coalescing for the NIC RX queues where busy polling will occur. This can help avoid the userland app from being pre-empted by a hard IRQ while userland is running. Note this means that userland must take care to call epoll_wait and not take too long in userland since it now drives NAPI via epoll_wait. 3. Optionally: Consider using napi_defer_hard_irqs and gro_flush_timeout to further restrict IRQ generation from the NIC. These settings are system-wide so their impact must be carefully weighed against the running applications. 4. Ensure that all incoming connections added to an epoll instance have the same NAPI ID. This can be done with a BPF filter when SO_REUSEPORT is used or getsockopt + SO_INCOMING_NAPI_ID when a single accept thread is used which dispatches incoming connections to threads. 5. Lastly, busy poll must be enabled via a sysctl (/proc/sys/net/core/busy_poll). Please see Eric Dumazet's paper about busy polling [3] and a recent academic paper about measured performance improvements of busy polling [4] (albeit with a modification that is not currently present in the kernel) for additional context. The unfortunate part about step 5 above is that this enables busy poll system-wide which affects all user applications on the system, including epoll-based network applications which were not intended to be used this way or applications where increased CPU usage for lower latency network processing is unnecessary or not desirable. If the user wants to run one low latency epoll-based server application with epoll-based busy poll, but would like to run the rest of the applications on the system (which may also use epoll) without busy poll, this system-wide sysctl presents a significant problem. This change preserves the system-wide sysctl, but adds a mechanism (via ioctl) to enable or disable busy poll for epoll contexts as needed by individual applications, making epoll-based busy poll more usable. Note that this change includes an or (as of v4) instead of an xor. If the user has enabled both the system-wide sysctl and also the per epoll-context busy poll settings, then epoll should probably busy poll (vs being disabled). Thanks, Joe v7 -> v8: - Reviewed-by tag from Eric Dumazet applied to commit message of patch 1/4. - patch 4/4: - EPIOCSPARAMS and EPIOCGPARAMS updated to use WRITE_ONCE and READ_ONCE, as requested by Eric Dumazet - Wrapped a long line (via netdev/checkpatch) v6 -> v7: - Acked-by tags from Stanislav Fomichev applied to commit messages of all patches. - Reviewed-by tags from Jakub Kicinski, Eric Dumazet applied to commit messages of patches 2 and 3. Jiri Slaby's Reviewed-by applied to patch 4. - patch 1/4: - busy_poll_usecs reduced from u64 to u32. - Unnecessary parens removed (via netdev/checkpatch) - Wrapped long line (via netdev/checkpatch) - Remove inline from busy_loop_ep_timeout as objdump suggests the function is already inlined - Moved struct eventpoll assignment to declaration - busy_loop_ep_timeout is moved within CONFIG_NET_RX_BUSY_POLL and the ifdefs internally have been removed as per Eric Dumazet's review - Removed ep_busy_loop_on from the !defined CONFIG_NET_RX_BUSY_POLL section as it is only called when CONFIG_NET_RX_BUSY_POLL is defined - patch 3/4: - Fix whitespace alignment issue (via netdev/checkpatch) - patch 4/4: - epoll_params.busy_poll_usecs has been reduced to u32 - epoll_params.busy_poll_usecs is now checked to ensure it is <= S32_MAX - __pad has been reduced to a single u8 - memchr_inv has been dropped and replaced with a simple check for the single __pad byte - Removed space after cast (via netdev/checkpatch) - Wrap long line (via netdev/checkpatch) - Move struct eventpoll *ep assignment to declaration as per Jiri Slaby's review - Remove unnecessary !! as per Jiri Slaby's review - Reorganized variables to be reverse christmas tree order v5 -> v6: - patch 1/3 no functional change, but commit message corrected to explain that an or (||) is being used instead of xor. - patch 3/4 is a new patch which adds support for per epoll context prefer busy poll setting. - patch 4/4 updated to allow getting/setting per epoll context prefer busy poll setting; this setting is limited to either 0 or 1. v4 -> v5: - patch 3/3 updated to use memchr_inv to ensure that __pad is zero for the EPIOCSPARAMS ioctl. Recommended by Greg K-H [5], Dave Chinner [6], and Jiri Slaby [7]. v3 -> v4: - patch 1/3 was updated to include an important functional change: ep_busy_loop_on was updated to use or (||) instead of xor (^). After thinking about it a bit more, I thought xor didn't make much sense. Enabling both the per-epoll context and the system-wide sysctl should probably enable busy poll, not disable it. So, or (||) makes more sense, I think. - patch 3/3 was updated: - to change the epoll_params fields to be __u64, __u16, and __u8 and to pad the struct to a multiple of 64bits. Suggested by Greg K-H [8] and Arnd Bergmann [9]. - remove an unused pr_fmt, left over from the previous revision. - ioctl now returns -EINVAL when epoll_params.busy_poll_usecs > U32_MAX. v2 -> v3: - cover letter updated to mention why ioctl seems (to me) like a better choice vs a new syscall. - patch 3/4 was modified in 3 ways: - when an unknown ioctl is received, -ENOIOCTLCMD is returned instead of -EINVAL as the ioctl documentation requires. - epoll_params.busy_poll_budget can only be set to a value larger than NAPI_POLL_WEIGHT if code is run by privileged (CAP_NET_ADMIN) users. Otherwise, -EPERM is returned. - busy poll specific ioctl code moved out to its own function. On kernels without busy poll support, -EOPNOTSUPP is returned. This also makes the kernel build robot happier without littering the code with more #ifdefs. - dropped patch 4/4 after Eric Dumazet's review of it when it was sent independently to the list [10]. v1 -> v2: - cover letter updated to make a mention of napi_defer_hard_irqs and gro_flush_timeout as an added step 3 and to cite both Eric Dumazet's busy polling paper and a paper from University of Waterloo for additional context. Specifically calling out the xor in patch 1/4 incase it is missed by reviewers. - Patch 2/4 has its commit message updated, but no functional changes. Commit message now describes that allowing for a settable budget helps to improve throughput and is more consistent with other busy poll mechanisms that allow a settable budget via SO_BUSY_POLL_BUDGET. - Patch 3/4 was modified to check if the epoll_params.busy_poll_budget exceeds NAPI_POLL_WEIGHT. The larger value is allowed, but an error is printed. This was done for consistency with netif_napi_add_weight, which does the same. - Patch 3/4 the struct epoll_params was updated to fix the type of the data field; it was uint8_t and was changed to u8. - Patch 4/4 added to check if SO_BUSY_POLL_BUDGET exceeds NAPI_POLL_WEIGHT. The larger value is allowed, but an error is printed. This was done for consistency with netif_napi_add_weight, which does the same. ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
Showing
Please register or sign in to comment