From nobody Mon Feb 9 00:25:50 2026 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 1F2C67FBAA for ; Wed, 21 Feb 2024 14:06:41 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1708524401; cv=none; b=h7QBm2EHYuZRDNDfck7mVTYKPFkBXmP22ST/CkyAfjvUY4PKwm83LW0ei8TkyopTYX7kWSyDtI9eR46mHUvPchd5UiqU3y/ItqKZ5lkKtc/BvzIALzI9oORDSqVNJh0KJ1umRqVv9zfiqMK1SM5d/yXYPXK0WppS09uMQoNBohU= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1708524401; c=relaxed/simple; bh=fm2eB8i9EuVFNMUufsa6D8T3dp7XGQNhLUPFoO/jjZA=; h=Message-ID:Date:From:To:Cc:Subject:References:MIME-Version: Content-Type; b=lmCGZF6Ss2VDHihMaJrp8JqmvSvb5cNEOhL5G/Z2fWE3Y2oi3Pp4EONuMX1qeLFz0NRGabiGM91FteMnh/9pN/tqOjwalE2rJ+y13DrNPtOgWZYXtbcEjAnuyPWRjYi8/Mjy5EhZisfYUYnnOEdI7PhgOHYrqZemdNtVSSIa7HQ= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; arc=none smtp.client-ip=10.30.226.201 Received: by smtp.kernel.org (Postfix) with ESMTPSA id DC0DAC43390; Wed, 21 Feb 2024 14:06:40 +0000 (UTC) Received: from rostedt by gandalf with local (Exim 4.97) (envelope-from ) id 1rcnGv-00000002iAK-0W7f; Wed, 21 Feb 2024 09:08:29 -0500 Message-ID: <20240221140828.986923313@goodmis.org> User-Agent: quilt/0.67 Date: Wed, 21 Feb 2024 09:08:03 -0500 From: Steven Rostedt To: linux-kernel@vger.kernel.org Cc: Masami Hiramatsu , Mark Rutland , Mathieu Desnoyers , Andrew Morton , Vincent Donnefort Subject: [for-next][PATCH 07/11] ring-buffer: Introducing ring-buffer mapping functions References: <20240221140756.797572998@goodmis.org> Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" From: Vincent Donnefort In preparation for allowing the user-space to map a ring-buffer, add a set of mapping functions: ring_buffer_{map,unmap}() ring_buffer_map_fault() And controls on the ring-buffer: ring_buffer_map_get_reader() /* swap reader and head */ Mapping the ring-buffer also involves: A unique ID for each subbuf of the ring-buffer, currently they are only identified through their in-kernel VA. A meta-page, where are stored ring-buffer statistics and a description for the current reader The linear mapping exposes the meta-page, and each subbuf of the ring-buffer, ordered following their unique ID, assigned during the first mapping. Once mapped, no subbuf can get in or out of the ring-buffer: the buffer size will remain unmodified and the splice enabling functions will in reality simply memcpy the data instead of swapping subbufs. Link: https://lore.kernel.org/linux-trace-kernel/20240220202310.2489614-3-v= donnefort@google.com Signed-off-by: Vincent Donnefort Signed-off-by: Steven Rostedt (Google) --- include/linux/ring_buffer.h | 7 + include/uapi/linux/trace_mmap.h | 46 ++++ kernel/trace/ring_buffer.c | 376 +++++++++++++++++++++++++++++++- 3 files changed, 426 insertions(+), 3 deletions(-) create mode 100644 include/uapi/linux/trace_mmap.h diff --git a/include/linux/ring_buffer.h b/include/linux/ring_buffer.h index fa802db216f9..0841ba8bab14 100644 --- a/include/linux/ring_buffer.h +++ b/include/linux/ring_buffer.h @@ -6,6 +6,8 @@ #include #include =20 +#include + struct trace_buffer; struct ring_buffer_iter; =20 @@ -221,4 +223,9 @@ int trace_rb_cpu_prepare(unsigned int cpu, struct hlist= _node *node); #define trace_rb_cpu_prepare NULL #endif =20 +int ring_buffer_map(struct trace_buffer *buffer, int cpu); +int ring_buffer_unmap(struct trace_buffer *buffer, int cpu); +struct page *ring_buffer_map_fault(struct trace_buffer *buffer, int cpu, + unsigned long pgoff); +int ring_buffer_map_get_reader(struct trace_buffer *buffer, int cpu); #endif /* _LINUX_RING_BUFFER_H */ diff --git a/include/uapi/linux/trace_mmap.h b/include/uapi/linux/trace_mma= p.h new file mode 100644 index 000000000000..ffcd8dfcaa4f --- /dev/null +++ b/include/uapi/linux/trace_mmap.h @@ -0,0 +1,46 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _TRACE_MMAP_H_ +#define _TRACE_MMAP_H_ + +#include + +/** + * struct trace_buffer_meta - Ring-buffer Meta-page description + * @meta_page_size: Size of this meta-page. + * @meta_struct_len: Size of this structure. + * @subbuf_size: Size of each sub-buffer. + * @nr_subbufs: Number of subbfs in the ring-buffer, including the reader. + * @reader.lost_events: Number of events lost at the time of the reader sw= ap. + * @reader.id: subbuf ID of the current reader. ID range [0 : @nr_subbufs= - 1] + * @reader.read: Number of bytes read on the reader subbuf. + * @flags: Placeholder for now, 0 until new features are supported. + * @entries: Number of entries in the ring-buffer. + * @overrun: Number of entries lost in the ring-buffer. + * @read: Number of entries that have been read. + * @Reserved1: Reserved for future use. + * @Reserved2: Reserved for future use. + */ +struct trace_buffer_meta { + __u32 meta_page_size; + __u32 meta_struct_len; + + __u32 subbuf_size; + __u32 nr_subbufs; + + struct { + __u64 lost_events; + __u32 id; + __u32 read; + } reader; + + __u64 flags; + + __u64 entries; + __u64 overrun; + __u64 read; + + __u64 Reserved1; + __u64 Reserved2; +}; + +#endif /* _TRACE_MMAP_H_ */ diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index ca796675c0a1..1d7d7a701867 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -9,6 +9,7 @@ #include #include #include +#include #include #include #include @@ -338,6 +339,7 @@ struct buffer_page { local_t entries; /* entries on this page */ unsigned long real_end; /* real end of data */ unsigned order; /* order of the page */ + u32 id; /* ID for external mapping */ struct buffer_data_page *page; /* Actual data page */ }; =20 @@ -484,6 +486,12 @@ struct ring_buffer_per_cpu { u64 read_stamp; /* pages removed since last reset */ unsigned long pages_removed; + + unsigned int mapped; + struct mutex mapping_lock; + unsigned long *subbuf_ids; /* ID to subbuf addr */ + struct trace_buffer_meta *meta_page; + /* ring buffer pages to update, > 0 to add, < 0 to remove */ long nr_pages_to_update; struct list_head new_pages; /* new pages to add */ @@ -1548,6 +1556,7 @@ rb_allocate_cpu_buffer(struct trace_buffer *buffer, l= ong nr_pages, int cpu) init_irq_work(&cpu_buffer->irq_work.work, rb_wake_up_waiters); init_waitqueue_head(&cpu_buffer->irq_work.waiters); init_waitqueue_head(&cpu_buffer->irq_work.full_waiters); + mutex_init(&cpu_buffer->mapping_lock); =20 bpage =3D kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), GFP_KERNEL, cpu_to_node(cpu)); @@ -1738,8 +1747,6 @@ bool ring_buffer_time_stamp_abs(struct trace_buffer *= buffer) return buffer->time_stamp_abs; } =20 -static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer); - static inline unsigned long rb_page_entries(struct buffer_page *bpage) { return local_read(&bpage->entries) & RB_WRITE_MASK; @@ -5160,6 +5167,22 @@ static void rb_clear_buffer_page(struct buffer_page = *page) page->read =3D 0; } =20 +static void rb_update_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + struct trace_buffer_meta *meta =3D cpu_buffer->meta_page; + + meta->reader.read =3D cpu_buffer->reader_page->read; + meta->reader.id =3D cpu_buffer->reader_page->id; + meta->reader.lost_events =3D cpu_buffer->lost_events; + + meta->entries =3D local_read(&cpu_buffer->entries); + meta->overrun =3D local_read(&cpu_buffer->overrun); + meta->read =3D cpu_buffer->read; + + /* Some archs do not have data cache coherency between kernel and user-sp= ace */ + flush_dcache_folio(virt_to_folio(cpu_buffer->meta_page)); +} + static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) { @@ -5204,6 +5227,9 @@ rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) cpu_buffer->lost_events =3D 0; cpu_buffer->last_overrun =3D 0; =20 + if (cpu_buffer->mapped) + rb_update_meta_page(cpu_buffer); + rb_head_page_activate(cpu_buffer); cpu_buffer->pages_removed =3D 0; } @@ -5418,6 +5444,12 @@ int ring_buffer_swap_cpu(struct trace_buffer *buffer= _a, cpu_buffer_a =3D buffer_a->buffers[cpu]; cpu_buffer_b =3D buffer_b->buffers[cpu]; =20 + /* It's up to the callers to not try to swap mapped buffers */ + if (WARN_ON_ONCE(cpu_buffer_a->mapped || cpu_buffer_b->mapped)) { + ret =3D -EBUSY; + goto out; + } + /* At least make sure the two buffers are somewhat the same */ if (cpu_buffer_a->nr_pages !=3D cpu_buffer_b->nr_pages) goto out; @@ -5682,7 +5714,8 @@ int ring_buffer_read_page(struct trace_buffer *buffer, * Otherwise, we can simply swap the page with the one passed in. */ if (read || (len < (commit - read)) || - cpu_buffer->reader_page =3D=3D cpu_buffer->commit_page) { + cpu_buffer->reader_page =3D=3D cpu_buffer->commit_page || + cpu_buffer->mapped) { struct buffer_data_page *rpage =3D cpu_buffer->reader_page->page; unsigned int rpos =3D read; unsigned int pos =3D 0; @@ -5901,6 +5934,11 @@ int ring_buffer_subbuf_order_set(struct trace_buffer= *buffer, int order) =20 cpu_buffer =3D buffer->buffers[cpu]; =20 + if (cpu_buffer->mapped) { + err =3D -EBUSY; + goto error; + } + /* Update the number of pages to match the new size */ nr_pages =3D old_size * buffer->buffers[cpu]->nr_pages; nr_pages =3D DIV_ROUND_UP(nr_pages, buffer->subbuf_size); @@ -6002,6 +6040,338 @@ int ring_buffer_subbuf_order_set(struct trace_buffe= r *buffer, int order) } EXPORT_SYMBOL_GPL(ring_buffer_subbuf_order_set); =20 +#define subbuf_page(off, start) \ + virt_to_page((void *)(start + (off << PAGE_SHIFT))) + +#define foreach_subbuf_page(sub_order, start, page) \ + page =3D subbuf_page(0, (start)); \ + for (int __off =3D 0; __off < (1 << (sub_order)); \ + __off++, page =3D subbuf_page(__off, (start))) + +static inline void subbuf_map_prepare(unsigned long subbuf_start, int orde= r) +{ + struct page *page; + + /* + * When allocating order > 0 pages, only the first struct page has a + * refcount > 1. Increasing the refcount here ensures none of the struct + * page composing the sub-buffer is freeed when the mapping is closed. + */ + foreach_subbuf_page(order, subbuf_start, page) + page_ref_inc(page); +} + +static inline void subbuf_unmap(unsigned long subbuf_start, int order) +{ + struct page *page; + + foreach_subbuf_page(order, subbuf_start, page) { + page_ref_dec(page); + page->mapping =3D NULL; + } +} + +static void rb_free_subbuf_ids(struct ring_buffer_per_cpu *cpu_buffer) +{ + int sub_id; + + for (sub_id =3D 0; sub_id < cpu_buffer->nr_pages + 1; sub_id++) + subbuf_unmap(cpu_buffer->subbuf_ids[sub_id], + cpu_buffer->buffer->subbuf_order); + + kfree(cpu_buffer->subbuf_ids); + cpu_buffer->subbuf_ids =3D NULL; +} + +static int rb_alloc_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + struct page *page; + + if (cpu_buffer->meta_page) + return 0; + + page =3D alloc_page(GFP_USER | __GFP_ZERO); + if (!page) + return -ENOMEM; + + cpu_buffer->meta_page =3D page_to_virt(page); + + return 0; +} + +static void rb_free_meta_page(struct ring_buffer_per_cpu *cpu_buffer) +{ + unsigned long addr =3D (unsigned long)cpu_buffer->meta_page; + + if (!addr) + return; + + virt_to_page((void *)addr)->mapping =3D NULL; + free_page(addr); + cpu_buffer->meta_page =3D NULL; +} + +static void rb_setup_ids_meta_page(struct ring_buffer_per_cpu *cpu_buffer, + unsigned long *subbuf_ids) +{ + struct trace_buffer_meta *meta =3D cpu_buffer->meta_page; + unsigned int nr_subbufs =3D cpu_buffer->nr_pages + 1; + struct buffer_page *first_subbuf, *subbuf; + int id =3D 0; + + subbuf_ids[id] =3D (unsigned long)cpu_buffer->reader_page->page; + subbuf_map_prepare(subbuf_ids[id], cpu_buffer->buffer->subbuf_order); + cpu_buffer->reader_page->id =3D id++; + + first_subbuf =3D subbuf =3D rb_set_head_page(cpu_buffer); + do { + if (WARN_ON(id >=3D nr_subbufs)) + break; + + subbuf_ids[id] =3D (unsigned long)subbuf->page; + subbuf->id =3D id; + subbuf_map_prepare(subbuf_ids[id], cpu_buffer->buffer->subbuf_order); + + rb_inc_page(&subbuf); + id++; + } while (subbuf !=3D first_subbuf); + + /* install subbuf ID to kern VA translation */ + cpu_buffer->subbuf_ids =3D subbuf_ids; + + meta->meta_page_size =3D PAGE_SIZE; + meta->meta_struct_len =3D sizeof(*meta); + meta->nr_subbufs =3D nr_subbufs; + meta->subbuf_size =3D cpu_buffer->buffer->subbuf_size + BUF_PAGE_HDR_SIZE; + + rb_update_meta_page(cpu_buffer); +} + +static inline struct ring_buffer_per_cpu * +rb_get_mapped_buffer(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + + if (!cpumask_test_cpu(cpu, buffer->cpumask)) + return ERR_PTR(-EINVAL); + + cpu_buffer =3D buffer->buffers[cpu]; + + mutex_lock(&cpu_buffer->mapping_lock); + + if (!cpu_buffer->mapped) { + mutex_unlock(&cpu_buffer->mapping_lock); + return ERR_PTR(-ENODEV); + } + + return cpu_buffer; +} + +static inline void rb_put_mapped_buffer(struct ring_buffer_per_cpu *cpu_bu= ffer) +{ + mutex_unlock(&cpu_buffer->mapping_lock); +} + +/* + * Fast-path for rb_buffer_(un)map(). Called whenever the meta-page doesn'= t need + * to be set-up or torn-down. + */ +static int __rb_inc_dec_mapped(struct trace_buffer *buffer, + struct ring_buffer_per_cpu *cpu_buffer, + bool inc) +{ + unsigned long flags; + + lockdep_assert_held(&cpu_buffer->mapping_lock); + + if (inc && cpu_buffer->mapped =3D=3D UINT_MAX) + return -EBUSY; + + if (WARN_ON(!inc && cpu_buffer->mapped =3D=3D 0)) + return -EINVAL; + + mutex_lock(&buffer->mutex); + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); + + if (inc) + cpu_buffer->mapped++; + else + cpu_buffer->mapped--; + + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); + mutex_unlock(&buffer->mutex); + + return 0; +} + +int ring_buffer_map(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + unsigned long flags, *subbuf_ids; + int err =3D 0; + + if (!cpumask_test_cpu(cpu, buffer->cpumask)) + return -EINVAL; + + cpu_buffer =3D buffer->buffers[cpu]; + + mutex_lock(&cpu_buffer->mapping_lock); + + if (cpu_buffer->mapped) { + err =3D __rb_inc_dec_mapped(buffer, cpu_buffer, true); + mutex_unlock(&cpu_buffer->mapping_lock); + return err; + } + + /* prevent another thread from changing buffer/sub-buffer sizes */ + mutex_lock(&buffer->mutex); + + err =3D rb_alloc_meta_page(cpu_buffer); + if (err) + goto unlock; + + /* subbuf_ids include the reader while nr_pages does not */ + subbuf_ids =3D kcalloc(cpu_buffer->nr_pages + 1, sizeof(*subbuf_ids), GFP= _KERNEL); + if (!subbuf_ids) { + rb_free_meta_page(cpu_buffer); + err =3D -ENOMEM; + goto unlock; + } + + atomic_inc(&cpu_buffer->resize_disabled); + + /* + * Lock all readers to block any subbuf swap until the subbuf IDs are + * assigned. + */ + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); + + rb_setup_ids_meta_page(cpu_buffer, subbuf_ids); + cpu_buffer->mapped =3D 1; + + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); +unlock: + mutex_unlock(&buffer->mutex); + mutex_unlock(&cpu_buffer->mapping_lock); + + return err; +} + +int ring_buffer_unmap(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + unsigned long flags; + int err =3D 0; + + if (!cpumask_test_cpu(cpu, buffer->cpumask)) + return -EINVAL; + + cpu_buffer =3D buffer->buffers[cpu]; + + mutex_lock(&cpu_buffer->mapping_lock); + + if (!cpu_buffer->mapped) { + err =3D -ENODEV; + goto out; + } else if (cpu_buffer->mapped > 1) { + __rb_inc_dec_mapped(buffer, cpu_buffer, false); + goto out; + } + + mutex_lock(&buffer->mutex); + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); + + cpu_buffer->mapped =3D 0; + + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); + + rb_free_subbuf_ids(cpu_buffer); + rb_free_meta_page(cpu_buffer); + atomic_dec(&cpu_buffer->resize_disabled); + + mutex_unlock(&buffer->mutex); +out: + mutex_unlock(&cpu_buffer->mapping_lock); + + return err; +} + +/* + * +--------------+ pgoff =3D=3D 0 + * | meta page | + * +--------------+ pgoff =3D=3D 1 + * | subbuffer 0 | + * +--------------+ pgoff =3D=3D 1 + (1 << subbuf_order) + * | subbuffer 1 | + * ... + */ +struct page *ring_buffer_map_fault(struct trace_buffer *buffer, int cpu, + unsigned long pgoff) +{ + struct ring_buffer_per_cpu *cpu_buffer =3D buffer->buffers[cpu]; + unsigned long subbuf_id, subbuf_offset, addr; + struct page *page; + + if (!pgoff) + return virt_to_page((void *)cpu_buffer->meta_page); + + pgoff--; + + subbuf_id =3D pgoff >> buffer->subbuf_order; + if (subbuf_id > cpu_buffer->nr_pages) + return NULL; + + subbuf_offset =3D pgoff & ((1UL << buffer->subbuf_order) - 1); + addr =3D cpu_buffer->subbuf_ids[subbuf_id] + (subbuf_offset * PAGE_SIZE); + page =3D virt_to_page((void *)addr); + + return page; +} + +int ring_buffer_map_get_reader(struct trace_buffer *buffer, int cpu) +{ + struct ring_buffer_per_cpu *cpu_buffer; + unsigned long reader_size; + unsigned long flags; + + cpu_buffer =3D rb_get_mapped_buffer(buffer, cpu); + if (IS_ERR(cpu_buffer)) + return (int)PTR_ERR(cpu_buffer); + + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); +consume: + if (rb_per_cpu_empty(cpu_buffer)) + goto out; + + reader_size =3D rb_page_size(cpu_buffer->reader_page); + + /* + * There are data to be read on the current reader page, we can + * return to the caller. But before that, we assume the latter will read + * everything. Let's update the kernel reader accordingly. + */ + if (cpu_buffer->reader_page->read < reader_size) { + while (cpu_buffer->reader_page->read < reader_size) + rb_advance_reader(cpu_buffer); + goto out; + } + + if (WARN_ON(!rb_get_reader_page(cpu_buffer))) + goto out; + + goto consume; +out: + /* Some archs do not have data cache coherency between kernel and user-sp= ace */ + flush_dcache_folio(virt_to_folio(cpu_buffer->reader_page->page)); + + rb_update_meta_page(cpu_buffer); + + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); + rb_put_mapped_buffer(cpu_buffer); + + return 0; +} + /* * We only allocate new buffers, never free them if the CPU goes down. * If we were to free the buffer, then the user would lose any trace that = was in --=20 2.43.0