From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-10.1 required=3.0 tests=DKIMWL_WL_HIGH,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,INCLUDES_PATCH,MAILING_LIST_MULTI,SPF_PASS, USER_AGENT_MUTT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 43101C43387 for ; Fri, 11 Jan 2019 15:55:44 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 092C721848 for ; Fri, 11 Jan 2019 15:55:44 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=default; t=1547222144; bh=aZ+iU4NnSiExjNpf2JvFrRWoSzs3TMMIbdxypyKl51U=; h=Date:From:To:Cc:Subject:List-ID:From; b=BBTCM7837eG59W6/KPd6RNms6zMCa/vWO5gSPGxIRnuJSAQOjW3HEB21ZEEyk78iG OafYmM8+PXrUGhaEikgz+P8+Q5qDJyTgkDdzJXGV1wtWGcHomT6G2W1Mx5jiAI52Xi BfpGx1oBvfkZICfYj436HTTeItefWtXN+at0673E= Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1733148AbfAKPzm (ORCPT ); Fri, 11 Jan 2019 10:55:42 -0500 Received: from mail.kernel.org ([198.145.29.99]:45136 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1730850AbfAKPzm (ORCPT ); Fri, 11 Jan 2019 10:55:42 -0500 Received: from quaco.ghostprotocols.net (unknown [190.15.121.82]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPSA id E08CB20874; Fri, 11 Jan 2019 15:55:40 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=default; t=1547222141; bh=aZ+iU4NnSiExjNpf2JvFrRWoSzs3TMMIbdxypyKl51U=; h=Date:From:To:Cc:Subject:From; b=q2MbPgF8hsPhQWEy/nKCP1+OBYV7fss4oDsektPADkgdHDh5hn3wFfWzi8CJ+xGnG FShkV8Z/W7CecGWAmHsfEATmxf9H1lV/jtTLbd7fIWFf/8ahCmSCJCH/tkEV+tfkyj Vbq7enrgQXoIDy0aKR9XLSliowfMhYQnrI0xdNtA= Received: by quaco.ghostprotocols.net (Postfix, from userid 1000) id 8FFDD41AB5; Fri, 11 Jan 2019 12:55:38 -0300 (-03) Date: Fri, 11 Jan 2019 12:55:38 -0300 From: Arnaldo Carvalho de Melo To: Peter Zijlstra Cc: Ingo Molnar , Alexei Starovoitov , Daniel Borkmann , Jamal Hadi Salim , Linux Kernel Mailing List , Linux Networking Development Mailing List Subject: [PATCH/RFC] Make perf_event_open() propagate errors for use in bpf_perf_event_open() Message-ID: <20190111155538.GX22483@kernel.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-Url: http://acmel.wordpress.com User-Agent: Mutt/1.10.1 (2018-07-13) Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Peter, bpf_perf_event_open() already returns a value, but if perf_event_output's output_begin (mostly perf_output_begin) fails, the only way to know about that is looking before/after the rb->lost, right? For ring buffer users that is ok, we'll get a PERF_RECORD_LOST, etc, but for BPF programs it would be convenient to get that -ENOSPC and do some fallback, whatever makes sense, like in my augmented_syscalls stuff for 'perf trace', i.e. don't augment it (i.e. push stuff at the end of the normal payload), just don't filter the raw_syscalls:sys_enter, 'perf trace' will get the enter syscall enter event without the pointer dereference at the end, etc, warn the user but don't lose a syscall in the strace-like output. What do you think? Am I missing something? Probably ;-) Ah, its just test built. - Arnaldo diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 1d5c551a5add..9ed2af2abd6d 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -978,9 +978,9 @@ extern void perf_event_output_forward(struct perf_event *event, extern void perf_event_output_backward(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs); -extern void perf_event_output(struct perf_event *event, - struct perf_sample_data *data, - struct pt_regs *regs); +extern int perf_event_output(struct perf_event *event, + struct perf_sample_data *data, + struct pt_regs *regs); static inline bool is_default_overflow_handler(struct perf_event *event) diff --git a/kernel/events/core.c b/kernel/events/core.c index 3cd13a30f732..dcbb2b508034 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -6489,7 +6489,7 @@ void perf_prepare_sample(struct perf_event_header *header, data->phys_addr = perf_virt_to_phys(data->addr); } -static __always_inline void +static __always_inline int __perf_event_output(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs, @@ -6499,13 +6499,15 @@ __perf_event_output(struct perf_event *event, { struct perf_output_handle handle; struct perf_event_header header; + int err; /* protect the callchain buffers */ rcu_read_lock(); perf_prepare_sample(&header, data, event, regs); - if (output_begin(&handle, event, header.size)) + err = output_begin(&handle, event, header.size); + if (err) goto exit; perf_output_sample(&handle, &header, data, event); @@ -6514,6 +6516,7 @@ __perf_event_output(struct perf_event *event, exit: rcu_read_unlock(); + return err; } void @@ -6532,12 +6535,12 @@ perf_event_output_backward(struct perf_event *event, __perf_event_output(event, data, regs, perf_output_begin_backward); } -void +int perf_event_output(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { - __perf_event_output(event, data, regs, perf_output_begin); + return __perf_event_output(event, data, regs, perf_output_begin); } /* diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c index 8b068adb9da1..088c2032ceaf 100644 --- a/kernel/trace/bpf_trace.c +++ b/kernel/trace/bpf_trace.c @@ -431,8 +431,7 @@ __bpf_perf_event_output(struct pt_regs *regs, struct bpf_map *map, if (unlikely(event->oncpu != cpu)) return -EOPNOTSUPP; - perf_event_output(event, sd, regs); - return 0; + return perf_event_output(event, sd, regs); } BPF_CALL_5(bpf_perf_event_output, struct pt_regs *, regs, struct bpf_map *, map, diff --git a/tools/perf/examples/bpf/augmented_raw_syscalls.c b/tools/perf/examples/bpf/augmented_raw_syscalls.c index 53c233370fae..9e9d4c66e53c 100644 --- a/tools/perf/examples/bpf/augmented_raw_syscalls.c +++ b/tools/perf/examples/bpf/augmented_raw_syscalls.c @@ -141,8 +141,8 @@ int sys_enter(struct syscall_enter_args *args) len = sizeof(augmented_args.args); } - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, &augmented_args, len); - return 0; + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, &augmented_args, len); } SEC("raw_syscalls:sys_exit") diff --git a/tools/perf/examples/bpf/augmented_syscalls.c b/tools/perf/examples/bpf/augmented_syscalls.c index 2ae44813ef2d..b7dba114e36c 100644 --- a/tools/perf/examples/bpf/augmented_syscalls.c +++ b/tools/perf/examples/bpf/augmented_syscalls.c @@ -55,9 +55,9 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args) \ len -= sizeof(augmented_args.filename.value) - augmented_args.filename.size; \ len &= sizeof(augmented_args.filename.value) - 1; \ } \ - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ - &augmented_args, len); \ - return 0; \ + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ \ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ + &augmented_args, len); \ } \ int syscall_exit(syscall)(struct syscall_exit_args *args) \ { \ @@ -125,10 +125,10 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args) \ /* addrlen = augmented_args.args.addrlen; */ \ /* */ \ probe_read(&augmented_args.addr, addrlen, args->addr_ptr); \ - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ - &augmented_args, \ - sizeof(augmented_args) - sizeof(augmented_args.addr) + addrlen); \ - return 0; \ + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ \ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ + &augmented_args, \ + sizeof(augmented_args) - sizeof(augmented_args.addr) + addrlen);\ } \ int syscall_exit(syscall)(struct syscall_exit_args *args) \ { \ diff --git a/tools/perf/examples/bpf/etcsnoop.c b/tools/perf/examples/bpf/etcsnoop.c index b59e8812ee8c..550e69c2e8d1 100644 --- a/tools/perf/examples/bpf/etcsnoop.c +++ b/tools/perf/examples/bpf/etcsnoop.c @@ -49,11 +49,11 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args) \ args->filename_ptr); \ if (__builtin_memcmp(augmented_args.filename.value, etc, 4) != 0) \ return 0; \ - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ - &augmented_args, \ - (sizeof(augmented_args) - sizeof(augmented_args.filename.value) + \ - augmented_args.filename.size)); \ - return 0; \ + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ \ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ + &augmented_args, \ + (sizeof(augmented_args) - sizeof(augmented_args.filename.value) + \ + augmented_args.filename.size)); \ } struct syscall_enter_openat_args { From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnaldo Carvalho de Melo Subject: [PATCH/RFC] Make perf_event_open() propagate errors for use in bpf_perf_event_open() Date: Fri, 11 Jan 2019 12:55:38 -0300 Message-ID: <20190111155538.GX22483@kernel.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Ingo Molnar , Alexei Starovoitov , Daniel Borkmann , Jamal Hadi Salim , Linux Kernel Mailing List , Linux Networking Development Mailing List To: Peter Zijlstra Return-path: Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Hi Peter, bpf_perf_event_open() already returns a value, but if perf_event_output's output_begin (mostly perf_output_begin) fails, the only way to know about that is looking before/after the rb->lost, right? For ring buffer users that is ok, we'll get a PERF_RECORD_LOST, etc, but for BPF programs it would be convenient to get that -ENOSPC and do some fallback, whatever makes sense, like in my augmented_syscalls stuff for 'perf trace', i.e. don't augment it (i.e. push stuff at the end of the normal payload), just don't filter the raw_syscalls:sys_enter, 'perf trace' will get the enter syscall enter event without the pointer dereference at the end, etc, warn the user but don't lose a syscall in the strace-like output. What do you think? Am I missing something? Probably ;-) Ah, its just test built. - Arnaldo diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 1d5c551a5add..9ed2af2abd6d 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -978,9 +978,9 @@ extern void perf_event_output_forward(struct perf_event *event, extern void perf_event_output_backward(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs); -extern void perf_event_output(struct perf_event *event, - struct perf_sample_data *data, - struct pt_regs *regs); +extern int perf_event_output(struct perf_event *event, + struct perf_sample_data *data, + struct pt_regs *regs); static inline bool is_default_overflow_handler(struct perf_event *event) diff --git a/kernel/events/core.c b/kernel/events/core.c index 3cd13a30f732..dcbb2b508034 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -6489,7 +6489,7 @@ void perf_prepare_sample(struct perf_event_header *header, data->phys_addr = perf_virt_to_phys(data->addr); } -static __always_inline void +static __always_inline int __perf_event_output(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs, @@ -6499,13 +6499,15 @@ __perf_event_output(struct perf_event *event, { struct perf_output_handle handle; struct perf_event_header header; + int err; /* protect the callchain buffers */ rcu_read_lock(); perf_prepare_sample(&header, data, event, regs); - if (output_begin(&handle, event, header.size)) + err = output_begin(&handle, event, header.size); + if (err) goto exit; perf_output_sample(&handle, &header, data, event); @@ -6514,6 +6516,7 @@ __perf_event_output(struct perf_event *event, exit: rcu_read_unlock(); + return err; } void @@ -6532,12 +6535,12 @@ perf_event_output_backward(struct perf_event *event, __perf_event_output(event, data, regs, perf_output_begin_backward); } -void +int perf_event_output(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { - __perf_event_output(event, data, regs, perf_output_begin); + return __perf_event_output(event, data, regs, perf_output_begin); } /* diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c index 8b068adb9da1..088c2032ceaf 100644 --- a/kernel/trace/bpf_trace.c +++ b/kernel/trace/bpf_trace.c @@ -431,8 +431,7 @@ __bpf_perf_event_output(struct pt_regs *regs, struct bpf_map *map, if (unlikely(event->oncpu != cpu)) return -EOPNOTSUPP; - perf_event_output(event, sd, regs); - return 0; + return perf_event_output(event, sd, regs); } BPF_CALL_5(bpf_perf_event_output, struct pt_regs *, regs, struct bpf_map *, map, diff --git a/tools/perf/examples/bpf/augmented_raw_syscalls.c b/tools/perf/examples/bpf/augmented_raw_syscalls.c index 53c233370fae..9e9d4c66e53c 100644 --- a/tools/perf/examples/bpf/augmented_raw_syscalls.c +++ b/tools/perf/examples/bpf/augmented_raw_syscalls.c @@ -141,8 +141,8 @@ int sys_enter(struct syscall_enter_args *args) len = sizeof(augmented_args.args); } - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, &augmented_args, len); - return 0; + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, &augmented_args, len); } SEC("raw_syscalls:sys_exit") diff --git a/tools/perf/examples/bpf/augmented_syscalls.c b/tools/perf/examples/bpf/augmented_syscalls.c index 2ae44813ef2d..b7dba114e36c 100644 --- a/tools/perf/examples/bpf/augmented_syscalls.c +++ b/tools/perf/examples/bpf/augmented_syscalls.c @@ -55,9 +55,9 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args) \ len -= sizeof(augmented_args.filename.value) - augmented_args.filename.size; \ len &= sizeof(augmented_args.filename.value) - 1; \ } \ - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ - &augmented_args, len); \ - return 0; \ + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ \ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ + &augmented_args, len); \ } \ int syscall_exit(syscall)(struct syscall_exit_args *args) \ { \ @@ -125,10 +125,10 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args) \ /* addrlen = augmented_args.args.addrlen; */ \ /* */ \ probe_read(&augmented_args.addr, addrlen, args->addr_ptr); \ - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ - &augmented_args, \ - sizeof(augmented_args) - sizeof(augmented_args.addr) + addrlen); \ - return 0; \ + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ \ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ + &augmented_args, \ + sizeof(augmented_args) - sizeof(augmented_args.addr) + addrlen);\ } \ int syscall_exit(syscall)(struct syscall_exit_args *args) \ { \ diff --git a/tools/perf/examples/bpf/etcsnoop.c b/tools/perf/examples/bpf/etcsnoop.c index b59e8812ee8c..550e69c2e8d1 100644 --- a/tools/perf/examples/bpf/etcsnoop.c +++ b/tools/perf/examples/bpf/etcsnoop.c @@ -49,11 +49,11 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args) \ args->filename_ptr); \ if (__builtin_memcmp(augmented_args.filename.value, etc, 4) != 0) \ return 0; \ - perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ - &augmented_args, \ - (sizeof(augmented_args) - sizeof(augmented_args.filename.value) + \ - augmented_args.filename.size)); \ - return 0; \ + /* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */ \ + return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, \ + &augmented_args, \ + (sizeof(augmented_args) - sizeof(augmented_args.filename.value) + \ + augmented_args.filename.size)); \ } struct syscall_enter_openat_args {