- 05 Sep, 2022 11 commits
-
-
Alexei Starovoitov authored
Convert dynamic allocations in percpu hash map from alloc_percpu() to bpf_mem_cache_alloc() from per-cpu bpf_mem_alloc. Since bpf_mem_alloc frees objects after RCU gp the call_rcu() is removed. pcpu_init_value() now needs to zero-fill per-cpu allocations, since dynamically allocated map elements are now similar to full prealloc, since alloc_percpu() is not called inline and the elements are reused in the freelist. Signed-off-by:
Alexei Starovoitov <ast@kernel.org> Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
Kumar Kartikeya Dwivedi <memxor@gmail.com> Acked-by:
Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/bpf/20220902211058.60789-12-alexei.starovoitov@gmail.com
-
Alexei Starovoitov authored
Extend bpf_mem_alloc to cache free list of fixed size per-cpu allocations. Once such cache is created bpf_mem_cache_alloc() will return per-cpu objects. bpf_mem_cache_free() will free them back into global per-cpu pool after observing RCU grace period. per-cpu flavor of bpf_mem_alloc is going to be used by per-cpu hash maps. The free list cache consists of tuples { llist_node, per-cpu pointer } Unlike alloc_percpu() that returns per-cpu pointer the bpf_mem_cache_alloc() returns a pointer to per-cpu pointer and bpf_mem_cache_free() expects to receive it back. Signed-off-by: -
Alexei Starovoitov authored
SLAB_TYPESAFE_BY_RCU makes kmem_caches non mergeable and slows down kmem_cache_destroy. All bpf_mem_cache are safe to share across different maps and programs. Convert SLAB_TYPESAFE_BY_RCU to batched call_rcu. This change solves the memory consumption issue, avoids kmem_cache_destroy latency and keeps bpf hash map performance the same. Signed-off-by:
-
Alexei Starovoitov authored
The same low/high watermarks for every bucket in bpf_mem_cache consume significant amount of memory. Preallocating 64 elements of 4096 bytes each in the free list is not efficient. Make low/high watermarks and batching value dependent on element size. This change brings significant memory savings. Signed-off-by:
-
Alexei Starovoitov authored
Doing call_rcu() million times a second becomes a bottle neck. Convert non-preallocated hash map from call_rcu to SLAB_TYPESAFE_BY_RCU. The rcu critical section is no longer observed for one htab element which makes non-preallocated hash map behave just like preallocated hash map. The map elements are released back to kernel memory after observing rcu critical section. This improves 'map_perf_test 4' performance from 100k events per second to 250k events per second. bpf_mem_alloc + percpu_counter + typesafe_by_rcu provide 10x performance boost to non-preallocated hash map and make it within few % of preallocated map while consuming fraction of memory. Signed-off-by:
-
Alexei Starovoitov authored
The atomic_inc/dec might cause extreme cache line bouncing when multiple cpus access the same bpf map. Based on specified max_entries for the hash map calculate when percpu_counter becomes faster than atomic_t and use it for such maps. For example samples/bpf/map_perf_test is using hash map with max_entries 1000. On a system with 16 cpus the 'map_perf_test 4' shows 14k events per second using atomic_t. On a system with 15 cpus it shows 100k events per second using percpu. map_perf_test is an extreme case where all cpus colliding on atomic_t which causes extreme cache bouncing. Note that the slow path of percpu_counter is 5k events per secound vs 14k for atomic, so the heuristic is necessary. See comment in the code why the heuristic is based on num_online_cpus(). Signed-off-by:
-
Alexei Starovoitov authored
Since bpf hash map was converted to use bpf_mem_alloc it is safe to use from tracing programs and in RT kernels. But per-cpu hash map is still using dynamic allocation for per-cpu map values, hence keep the warning for this map type. In the future alloc_percpu_gfp can be front-end-ed with bpf_mem_cache and this restriction will be completely lifted. perf_event (NMI) bpf programs have to use preallocated hash maps, because free_htab_elem() is using call_rcu which might crash if re-entered. Sleepable bpf programs have to use preallocated hash maps, because life time of the map elements is not protected by rcu_read_lock/unlock. This restriction can be lifted in the future as well. Signed-off-by:
-
Alexei Starovoitov authored
Make map_perf_test for preallocated and non-preallocated hash map spend more time inside bpf program to focus performance analysis on the speed of update/lookup/delete operations performed by bpf program. It makes 'perf report' of bpf_mem_alloc look like: 11.76% map_perf_test [k] _raw_spin_lock_irqsave 11.26% map_perf_test [k] htab_map_update_elem 9.70% map_perf_test [k] _raw_spin_lock 9.47% map_perf_test [k] htab_map_delete_elem 8.57% map_perf_test [k] memcpy_erms 5.58% map_perf_test [k] alloc_htab_elem 4.09% map_perf_test [k] __htab_map_lookup_elem 3.44% map_perf_test [k] syscall_exit_to_user_mode 3.13% map_perf_test [k] lookup_nulls_elem_raw 3.05% map_perf_test [k] migrate_enable 3.04% map_perf_test [k] memcmp 2.67% map_perf_test [k] unit_free 2.39% map_perf_test [k] lookup_elem_raw Reduce default iteration count as well to make 'map_perf_test' quick enough even on debug kernels. Signed-off-by:
-
Alexei Starovoitov authored
Make test_maps more stressful with more parallelism in update/delete/lookup/walk including different value sizes. Signed-off-by:
-
Alexei Starovoitov authored
Convert bpf hash map to use bpf memory allocator. Signed-off-by:
-
Alexei Starovoitov authored
Tracing BPF programs can attach to kprobe and fentry. Hence they run in unknown context where calling plain kmalloc() might not be safe. Front-end kmalloc() with minimal per-cpu cache of free elements. Refill this cache asynchronously from irq_work. BPF programs always run with migration disabled. It's safe to allocate from cache of the current cpu with irqs disabled. Free-ing is always done into bucket of the current cpu as well. irq_work trims extra free elements from buckets with kfree and refills them with kmalloc, so global kmalloc logic takes care of freeing objects allocated by one cpu and freed on another. struct bpf_mem_alloc supports two modes: - When size != 0 create kmem_cache and bpf_mem_cache for each cpu. This is typical bpf hash map use case when all elements have equal size. - When size == 0 allocate 11 bpf_mem_cache-s for each cpu, then rely on kmalloc/kfree. Max allocation size is 4096 in this case. This is bpf_dynptr and bpf_kptr use case. bpf_mem_alloc/bpf_mem_free are bpf specific 'wrappers' of kmalloc/kfree. bpf_mem_cache_alloc/bpf_mem_cache_free are 'wrappers' of kmem_cache_alloc/kmem_cache_free. The allocators are NMI-safe from bpf programs only. They are not NMI-safe in general. Signed-off-by:
-
- 03 Sep, 2022 18 commits
-
-
Alexei Starovoitov authored
Martin KaFai Lau says: ==================== From: Martin KaFai Lau <martin.lau@kernel.org> The earlier commits [0] removed duplicated code from bpf_setsockopt(). This series is to remove duplicated code from bpf_getsockopt(). Unlike the setsockopt() which had already changed to take the sockptr_t argument, the same has not been done to getsockopt(). This is the extra step being done in this series. [0]: https://lore.kernel.org/all/20220817061704.4174272-1-kafai@fb.com/ v2: - The previous v2 did not reach the list. It is a resend. - Add comments on bpf_getsockopt() should not free the saved_syn (Stanislav) - Explicitly null-terminate the tcp-cc name (Stanislav) ==================== Reviewed-by:
Stanislav Fomichev <sdf@google.com> Signed-off-by:
-
Martin KaFai Lau authored
This patch removes the __bpf_getsockopt() which directly reads the sk by using PTR_TO_BTF_ID. Instead, the test now directly uses the kernel bpf helper bpf_getsockopt() which supports all the required optname now. TCP_SAVE[D]_SYN and TCP_MAXSEG are not tested in a loop for all the hooks and sock_ops's cb. TCP_SAVE[D]_SYN only works in passive connection. TCP_MAXSEG only works when it is setsockopt before the connection is established and the getsockopt return value can only be tested after the connection is established. Signed-off-by:
Martin KaFai Lau <martin.lau@kernel.org> Link: https://lore.kernel.org/r/20220902002937.2896904-1-kafai@fb.comSigned-off-by:
-
Martin KaFai Lau authored
This patch changes bpf_getsockopt(SOL_IPV6) to reuse do_ipv6_getsockopt(). It removes the duplicated code from bpf_getsockopt(SOL_IPV6). This also makes bpf_getsockopt(SOL_IPV6) supporting the same set of optnames as in bpf_setsockopt(SOL_IPV6). In particular, this adds IPV6_AUTOFLOWLABEL support to bpf_getsockopt(SOL_IPV6). ipv6 could be compiled as a module. Like how other code solved it with stubs in ipv6_stubs.h, this patch adds the do_ipv6_getsockopt to the ipv6_bpf_stub. Signed-off-by:
-
Martin KaFai Lau authored
This patch changes bpf_getsockopt(SOL_IP) to reuse do_ip_getsockopt() and remove the duplicated code. Signed-off-by:
-
Martin KaFai Lau authored
This patch changes bpf_getsockopt(SOL_TCP) to reuse do_tcp_getsockopt(). It removes the duplicated code from bpf_getsockopt(SOL_TCP). Before this patch, there were some optnames available to bpf_setsockopt(SOL_TCP) but missing in bpf_getsockopt(SOL_TCP). For example, TCP_NODELAY, TCP_MAXSEG, TCP_KEEPIDLE, TCP_KEEPINTVL, and a few more. It surprises users from time to time. This patch automatically closes this gap without duplicating more code. bpf_getsockopt(TCP_SAVED_SYN) does not free the saved_syn, so it stays in sol_tcp_sockopt(). For string name value like TCP_CONGESTION, bpf expects it is always null terminated, so sol_tcp_sockopt() decrements optlen by one before calling do_tcp_getsockopt() and the 'if (optlen < saved_optlen) memset(..,0,..);' in __bpf_getsockopt() will always do a null termination. Signed-off-by:
-
Martin KaFai Lau authored
This patch changes bpf_getsockopt(SOL_SOCKET) to reuse sk_getsockopt(). It removes all duplicated code from bpf_getsockopt(SOL_SOCKET). Before this patch, there were some optnames available to bpf_setsockopt(SOL_SOCKET) but missing in bpf_getsockopt(SOL_SOCKET). It surprises users from time to time. For example, SO_REUSEADDR, SO_KEEPALIVE, SO_RCVLOWAT, and SO_MAX_PACING_RATE. This patch automatically closes this gap without duplicating more code. The only exception is SO_BINDTODEVICE because it needs to acquire a blocking lock. Thus, SO_BINDTODEVICE is not supported. Signed-off-by:
-
Martin KaFai Lau authored
This patch moves the "#ifdef CONFIG_XXX" check into the "if/else" statement itself. The change is done for the bpf_getsockopt() function only. It will make the latter patches easier to follow without the surrounding ifdef macro. Signed-off-by:
-
Martin KaFai Lau authored
Similar to the earlier patch that changes sk_getsockopt() to use sockopt_{lock,release}_sock() such that it can avoid taking the lock when called from bpf. This patch also changes do_ipv6_getsockopt() to use sockopt_{lock,release}_sock() such that bpf_getsockopt(SOL_IPV6) can reuse do_ipv6_getsockopt(). Although bpf_getsockopt(SOL_IPV6) currently does not support optname that requires lock_sock(), using sockopt_{lock,release}_sock() consistently across *_getsockopt() will make future optname addition harder to miss the sockopt_{lock,release}_sock() usage. eg. when adding new optname that requires a lock and the new optname is needed in bpf_getsockopt() also. Signed-off-by: -
Martin KaFai Lau authored
Similar to the earlier patch that changes sk_getsockopt() to take the sockptr_t argument . This patch also changes do_ipv6_getsockopt() to take the sockptr_t argument such that a latter patch can make bpf_getsockopt(SOL_IPV6) to reuse do_ipv6_getsockopt(). Note on the change in ip6_mc_msfget(). This function is to return an array of sockaddr_storage in optval. This function is shared between ipv6_get_msfilter() and compat_ipv6_get_msfilter(). However, the sockaddr_storage is stored at different offset of the optval because of the difference between group_filter and compat_group_filter. Thus, a new 'ss_offset' argument is added to ip6_mc_msfget(). Signed-off-by:
-
Martin KaFai Lau authored
Pass the len to the compat_ipv6_get_msfilter() instead of compat_ipv6_get_msfilter() getting it again from optlen. Its counter part ipv6_get_msfilter() is also taking the len from do_ipv6_getsockopt(). Signed-off-by:
-
Martin KaFai Lau authored
The 'unsigned int flags' argument is always 0, so it can be removed. Signed-off-by:
-
Martin KaFai Lau authored
Similar to the earlier commit that changed sk_setsockopt() to use sockopt_{lock,release}_sock() such that it can avoid taking lock when called from bpf. This patch also changes do_ip_getsockopt() to use sockopt_{lock,release}_sock() such that a latter patch can make bpf_getsockopt(SOL_IP) to reuse do_ip_getsockopt(). Signed-off-by: -
Martin KaFai Lau authored
Similar to the earlier patch that changes sk_getsockopt() to take the sockptr_t argument. This patch also changes do_ip_getsockopt() to take the sockptr_t argument such that a latter patch can make bpf_getsockopt(SOL_IP) to reuse do_ip_getsockopt(). Note on the change in ip_mc_gsfget(). This function is to return an array of sockaddr_storage in optval. This function is shared between ip_get_mcast_msfilter() and compat_ip_get_mcast_msfilter(). However, the sockaddr_storage is stored at different offset of the optval because of the difference between group_filter and compat_group_filter. Thus, a new 'ss_offset' argument is added to ip_mc_gsfget(). Signed-off-by:
-
Martin KaFai Lau authored
Similar to the earlier commit that changed sk_setsockopt() to use sockopt_{lock,release}_sock() such that it can avoid taking lock when called from bpf. This patch also changes do_tcp_getsockopt() to use sockopt_{lock,release}_sock() such that a latter patch can make bpf_getsockopt(SOL_TCP) to reuse do_tcp_getsockopt(). Signed-off-by: -
Martin KaFai Lau authored
Similar to the earlier patch that changes sk_getsockopt() to take the sockptr_t argument . This patch also changes do_tcp_getsockopt() to take the sockptr_t argument such that a latter patch can make bpf_getsockopt(SOL_TCP) to reuse do_tcp_getsockopt(). Signed-off-by:
-
Martin KaFai Lau authored
Similar to the earlier commit that changed sk_setsockopt() to use sockopt_{lock,release}_sock() such that it can avoid taking lock when called from bpf. This patch also changes sk_getsockopt() to use sockopt_{lock,release}_sock() such that a latter patch can make bpf_getsockopt(SOL_SOCKET) to reuse sk_getsockopt(). Only sk_get_filter() requires this change and it is used by the optname SO_GET_FILTER. The '.getname' implementations in sock->ops->getname() is not changed also since bpf does not always have the sk->sk_socket pointer and cannot support SO_PEERNAME. Signed-off-by: -
Martin KaFai Lau authored
This patch changes sk_getsockopt() to take the sockptr_t argument such that it can be used by bpf_getsockopt(SOL_SOCKET) in a latter patch. security_socket_getpeersec_stream() is not changed. It stays with the __user ptr (optval.user and optlen.user) to avoid changes to other security hooks. bpf_getsockopt(SOL_SOCKET) also does not support SO_PEERSEC. Signed-off-by:
-
Martin KaFai Lau authored
A latter patch refactors bpf_getsockopt(SOL_SOCKET) with the sock_getsockopt() to avoid code duplication and code drift between the two duplicates. The current sock_getsockopt() takes sock ptr as the argument. The very first thing of this function is to get back the sk ptr by 'sk = sock->sk'. bpf_getsockopt() could be called when the sk does not have the sock ptr created. Meaning sk->sk_socket is NULL. For example, when a passive tcp connection has just been established but has yet been accept()-ed. Thus, it cannot use the sock_getsockopt(sk->sk_socket) or else it will pass a NULL ptr. This patch moves all sock_getsockopt implementation to the newly added sk_getsockopt(). The new sk_getsockopt() takes a sk ptr and immediately gets the sock ptr by 'sock = sk->sk_socket' The existing sock_getsockopt(sock) is changed to call sk_getsockopt(sock->sk). All existing callers have both sock->sk and sk->sk_socket pointer. The latter patch will make bpf_getsockopt(SOL_SOCKET) call sk_getsockopt(sk) directly. The bpf_getsockopt(SOL_SOCKET) does not use the optnames that require sk->sk_socket, so it will be safe. Signed-off-by:
-
- 02 Sep, 2022 11 commits
-
-
Ian Rogers authored
The put lowers the reference count to 0 and frees ctx, reading it afterwards is invalid. Move the put after the uses and determine the last use by the reference count being 1. Fixes: 39e940d4 ("selftests/xsk: Destroy BPF resources only when ctx refcount drops to 0") Signed-off-by:
Ian Rogers <irogers@google.com> Signed-off-by:
Magnus Karlsson <magnus.karlsson@intel.com> Link: https://lore.kernel.org/bpf/20220901202645.1463552-1-irogers@google.com
-
Daniel Müller authored
BPF object files are, in a way, the final artifact produced as part of the ahead-of-time compilation process. That makes them somewhat special compared to "regular" object files, which are a intermediate build artifacts that can typically be removed safely. As such, it can make sense to name them differently to make it easier to spot this difference at a glance. Among others, libbpf-bootstrap [0] has established the extension .bpf.o for BPF object files. It seems reasonable to follow this example and establish the same denomination for selftest build artifacts. To that end, this change adjusts the corresponding part of the build system and the test programs loading BPF object files to work with .bpf.o files. [0] https://github.com/libbpf/libbpf-bootstrapSuggested-by:
Daniel Müller <deso@posteo.net> Signed-off-by:
-
Maciej Fijalkowski authored
Introduce new mode to xdpxceiver responsible for testing AF_XDP zero copy support of driver that serves underlying physical device. When setting up test suite, determine whether driver has ZC support or not by trying to bind XSK ZC socket to the interface. If it succeeded, interpret it as ZC support being in place and do softirq and busy poll tests for zero copy mode. Note that Rx dropped tests are skipped since ZC path is not touching rx_dropped stat at all. Signed-off-by:
Maciej Fijalkowski <maciej.fijalkowski@intel.com> Signed-off-by:
-
Maciej Fijalkowski authored
For single threaded poll tests call pthread_kill() from main thread so that we are sure worker thread has finished its job and it is possible to proceed with next test types from test suite. It was observed that on some platforms it takes a bit longer for worker thread to exit and next test case sees device as busy in this case. Signed-off-by:
-
Maciej Fijalkowski authored
Currently, architecture of xdpxceiver is designed strictly for conducting veth based tests. Veth pair is created together with a network namespace and one of the veth interfaces is moved to the mentioned netns. Then, separate threads for Tx and Rx are spawned which will utilize described setup. Infrastructure described in the paragraph above can not be used for testing AF_XDP support on physical devices. That testing will be conducted on a single network interface and same queue. Xskxceiver needs to be extended to distinguish between veth tests and physical interface tests. Since same iface/queue id pair will be used by both Tx/Rx threads for physical device testing, Tx thread, which happen to run after the Rx thread, is going to create XSK socket with shared umem flag. In order to track this setting throughout the lifetime of spawned threads, introduce 'shared_umem' boolean variable to struct ifobject and set it to true when xdpxceiver is run against physical device. In such case, UMEM size needs to be doubled, so half of it will be used by Rx thread and other half by Tx thread. For two step based test types, value of XSKMAP element under key 0 has to be updated as there is now another socket for the second step. Also, to avoid race conditions when destroying XSK resources, move this activity to the main thread after spawned Rx and Tx threads have finished its job. This way it is possible to gracefully remove shared umem without introducing synchronization mechanisms. To run xsk selftests suite on physical device, append "-i $IFACE" when invoking test_xsk.sh. For veth based tests, simply skip it. When "-i $IFACE" is in place, under the hood test_xsk.sh will use $IFACE for both interfaces supplied to xdpxceiver, which in turn will interpret that this execution of test suite is for a physical device. Note that currently this makes it possible only to test SKB and DRV mode (in case underlying device has native XDP support). ZC testing support is added in a later patch. Signed-off-by:
-
Maciej Fijalkowski authored
So that "enp240s0f0" or such name can be used against xskxceiver. While at it, also extend character count for netns name. Signed-off-by:
-
Maciej Fijalkowski authored
In order to prepare xdpxceiver for physical device testing, let us introduce default Rx pkt stream. Reason for doing it is that physical device testing will use a UMEM with a doubled size where half of it will be used by Tx and other half by Rx. This means that pkt addresses will differ for Tx and Rx streams. Rx thread will initialize the xsk_umem_info::base_addr that is added here so that pkt_set(), when working on Rx UMEM will add this offset and second half of UMEM space will be used. Note that currently base_addr is 0 on both sides. Future commit will do the mentioned initialization. Previously, veth based testing worked on separate UMEMs, so single default stream was fine. Signed-off-by:
-
Maciej Fijalkowski authored
Currently, xdpxceiver assumes that underlying device supports XDP in native mode - it is fine by now since tests can run only on a veth pair. Future commit is going to allow running test suite against physical devices, so let us query the device if it is capable of running XDP programs in native mode. This way xdpxceiver will not try to run TEST_MODE_DRV if device being tested is not supporting it. Signed-off-by:
-
Shmulik Ladkani authored
Get the tunnel flags in {ipv6}vxlan_get_tunnel_src and ensure they are aligned with tunnel params set at {ipv6}vxlan_set_tunnel_dst. Signed-off-by:Shmulik Ladkani <shmulik.ladkani@gmail.com> Signed-off-by:
-
Shmulik Ladkani authored
Existing 'bpf_skb_get_tunnel_key' extracts various tunnel parameters (id, ttl, tos, local and remote) but does not expose ip_tunnel_info's tun_flags to the BPF program. It makes sense to expose tun_flags to the BPF program. Assume for example multiple GRE tunnels maintained on a single GRE interface in collect_md mode. The program expects origins to initiate over GRE, however different origins use different GRE characteristics (e.g. some prefer to use GRE checksum, some do not; some pass a GRE key, some do not, etc..). A BPF program getting tun_flags can therefore remember the relevant flags (e.g. TUNNEL_CSUM, TUNNEL_SEQ...) for each initiating remote. In the reply path, the program can use 'bpf_skb_set_tunnel_key' in order to correctly reply to the remote, using similar characteristics, based on the stored tunnel flags. Introduce BPF_F_TUNINFO_FLAGS flag for bpf_skb_get_tunnel_key. If specified, 'bpf_tunnel_key->tunnel_flags' is set with the tun_flags. Decided to use the existing unused 'tunnel_ext' as the storage for the 'tunnel_flags' in order to avoid changing bpf_tunnel_key's layout. Also, the following has been considered during the design: 1. Convert the "interesting" internal TUNNEL_xxx flags back to BPF_F_yyy and place into the new 'tunnel_flags' field. This has 2 drawbacks: - The BPF_F_yyy flags are from *set_tunnel_key* enumeration space, e.g. BPF_F_ZERO_CSUM_TX. It is awkward that it is "returned" into tunnel_flags from a *get_tunnel_key* call. - Not all "interesting" TUNNEL_xxx flags can be mapped to existing BPF_F_yyy flags, and it doesn't make sense to create new BPF_F_yyy flags just for purposes of the returned tunnel_flags. 2. Place key.tun_flags into 'tunnel_flags' but mask them, keeping only "interesting" flags. That's ok, but the drawback is that what's "interesting" for my usecase might be limiting for other usecases. Therefore I decided to expose what's in key.tun_flags *as is*, which seems most flexible. The BPF user can just choose to ignore bits he's not interested in. The TUNNEL_xxx are also UAPI, so no harm exposing them back in the get_tunnel_key call. Signed-off-by: -
Shung-Hsi Yu authored
Commit a657182a ("bpf: Don't use tnum_range on array range checking for poke descriptors") has shown that using tnum_range() as argument to tnum_in() can lead to misleading code that looks like tight bound check when in fact the actual allowed range is much wider. Document such behavior to warn against its usage in general, and suggest some scenario where result can be trusted. Signed-off-by:
Shung-Hsi Yu <shung-hsi.yu@suse.com> Signed-off-by:
-
