bpf: Introduce bpf_obj_new
Introduce type safe memory allocator bpf_obj_new for BPF programs. The
kernel side kfunc is named bpf_obj_new_impl, as passing hidden arguments
to kfuncs still requires having them in prototype, unlike BPF helpers
which always take 5 arguments and have them checked using bpf_func_proto
in verifier, ignoring unset argument types.
Introduce __ign suffix to ignore a specific kfunc argument during type
checks, then use this to introduce support for passing type metadata to
the bpf_obj_new_impl kfunc.
The user passes BTF ID of the type it wants to allocates in program BTF,
the verifier then rewrites the first argument as the size of this type,
after performing some sanity checks (to ensure it exists and it is a
struct type).
The second argument is also fixed up and passed by the verifier. This is
the btf_struct_meta for the type being allocated. It would be needed
mostly for the offset array which is required for zero initializing
special fields while leaving the rest of storage in unitialized state.
It would also be needed in the next patch to perform proper destruction
of the object's special fields.
Under the hood, bpf_obj_new will call bpf_mem_alloc and bpf_mem_free,
using the any context BPF memory allocator introduced recently. To this
end, a global instance of the BPF memory allocator is initialized on
boot to be used for this purpose. This 'bpf_global_ma' serves all
allocations for bpf_obj_new. In the future, bpf_obj_new variants will
allow specifying a custom allocator.
Note that now that bpf_obj_new can be used to allocate objects that can
be linked to BPF linked list (when future linked list helpers are
available), we need to also free the elements using bpf_mem_free.
However, since the draining of elements is done outside the
bpf_spin_lock, we need to do migrate_disable around the call since
bpf_list_head_free can be called from map free path where migration is
enabled. Otherwise, when called from BPF programs migration is already
disabled.
A convenience macro is included in the bpf_experimental.h header to hide
over the ugly details of the implementation, leading to user code
looking similar to a language level extension which allocates and
constructs fields of a user type.
struct bar {
struct bpf_list_node node;
};
struct foo {
struct bpf_spin_lock lock;
struct bpf_list_head head __contains(bar, node);
};
void prog(void) {
struct foo *f;
f = bpf_obj_new(typeof(*f));
if (!f)
return;
...
}
A key piece of this story is still missing, i.e. the free function,
which will come in the next patch.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-14-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:03 +00:00
|
|
|
#ifndef __BPF_EXPERIMENTAL__
|
|
|
|
|
#define __BPF_EXPERIMENTAL__
|
|
|
|
|
|
|
|
|
|
#include <vmlinux.h>
|
|
|
|
|
#include <bpf/bpf_tracing.h>
|
|
|
|
|
#include <bpf/bpf_helpers.h>
|
|
|
|
|
#include <bpf/bpf_core_read.h>
|
|
|
|
|
|
2022-11-18 01:56:09 +00:00
|
|
|
#define __contains(name, node) __attribute__((btf_decl_tag("contains:" #name ":" #node)))
|
|
|
|
|
|
bpf: Introduce bpf_obj_new
Introduce type safe memory allocator bpf_obj_new for BPF programs. The
kernel side kfunc is named bpf_obj_new_impl, as passing hidden arguments
to kfuncs still requires having them in prototype, unlike BPF helpers
which always take 5 arguments and have them checked using bpf_func_proto
in verifier, ignoring unset argument types.
Introduce __ign suffix to ignore a specific kfunc argument during type
checks, then use this to introduce support for passing type metadata to
the bpf_obj_new_impl kfunc.
The user passes BTF ID of the type it wants to allocates in program BTF,
the verifier then rewrites the first argument as the size of this type,
after performing some sanity checks (to ensure it exists and it is a
struct type).
The second argument is also fixed up and passed by the verifier. This is
the btf_struct_meta for the type being allocated. It would be needed
mostly for the offset array which is required for zero initializing
special fields while leaving the rest of storage in unitialized state.
It would also be needed in the next patch to perform proper destruction
of the object's special fields.
Under the hood, bpf_obj_new will call bpf_mem_alloc and bpf_mem_free,
using the any context BPF memory allocator introduced recently. To this
end, a global instance of the BPF memory allocator is initialized on
boot to be used for this purpose. This 'bpf_global_ma' serves all
allocations for bpf_obj_new. In the future, bpf_obj_new variants will
allow specifying a custom allocator.
Note that now that bpf_obj_new can be used to allocate objects that can
be linked to BPF linked list (when future linked list helpers are
available), we need to also free the elements using bpf_mem_free.
However, since the draining of elements is done outside the
bpf_spin_lock, we need to do migrate_disable around the call since
bpf_list_head_free can be called from map free path where migration is
enabled. Otherwise, when called from BPF programs migration is already
disabled.
A convenience macro is included in the bpf_experimental.h header to hide
over the ugly details of the implementation, leading to user code
looking similar to a language level extension which allocates and
constructs fields of a user type.
struct bar {
struct bpf_list_node node;
};
struct foo {
struct bpf_spin_lock lock;
struct bpf_list_head head __contains(bar, node);
};
void prog(void) {
struct foo *f;
f = bpf_obj_new(typeof(*f));
if (!f)
return;
...
}
A key piece of this story is still missing, i.e. the free function,
which will come in the next patch.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-14-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:03 +00:00
|
|
|
/* Description
|
|
|
|
|
* Allocates an object of the type represented by 'local_type_id' in
|
|
|
|
|
* program BTF. User may use the bpf_core_type_id_local macro to pass the
|
|
|
|
|
* type ID of a struct in program BTF.
|
|
|
|
|
*
|
|
|
|
|
* The 'local_type_id' parameter must be a known constant.
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
* The 'meta' parameter is rewritten by the verifier, no need for BPF
|
|
|
|
|
* program to set it.
|
bpf: Introduce bpf_obj_new
Introduce type safe memory allocator bpf_obj_new for BPF programs. The
kernel side kfunc is named bpf_obj_new_impl, as passing hidden arguments
to kfuncs still requires having them in prototype, unlike BPF helpers
which always take 5 arguments and have them checked using bpf_func_proto
in verifier, ignoring unset argument types.
Introduce __ign suffix to ignore a specific kfunc argument during type
checks, then use this to introduce support for passing type metadata to
the bpf_obj_new_impl kfunc.
The user passes BTF ID of the type it wants to allocates in program BTF,
the verifier then rewrites the first argument as the size of this type,
after performing some sanity checks (to ensure it exists and it is a
struct type).
The second argument is also fixed up and passed by the verifier. This is
the btf_struct_meta for the type being allocated. It would be needed
mostly for the offset array which is required for zero initializing
special fields while leaving the rest of storage in unitialized state.
It would also be needed in the next patch to perform proper destruction
of the object's special fields.
Under the hood, bpf_obj_new will call bpf_mem_alloc and bpf_mem_free,
using the any context BPF memory allocator introduced recently. To this
end, a global instance of the BPF memory allocator is initialized on
boot to be used for this purpose. This 'bpf_global_ma' serves all
allocations for bpf_obj_new. In the future, bpf_obj_new variants will
allow specifying a custom allocator.
Note that now that bpf_obj_new can be used to allocate objects that can
be linked to BPF linked list (when future linked list helpers are
available), we need to also free the elements using bpf_mem_free.
However, since the draining of elements is done outside the
bpf_spin_lock, we need to do migrate_disable around the call since
bpf_list_head_free can be called from map free path where migration is
enabled. Otherwise, when called from BPF programs migration is already
disabled.
A convenience macro is included in the bpf_experimental.h header to hide
over the ugly details of the implementation, leading to user code
looking similar to a language level extension which allocates and
constructs fields of a user type.
struct bar {
struct bpf_list_node node;
};
struct foo {
struct bpf_spin_lock lock;
struct bpf_list_head head __contains(bar, node);
};
void prog(void) {
struct foo *f;
f = bpf_obj_new(typeof(*f));
if (!f)
return;
...
}
A key piece of this story is still missing, i.e. the free function,
which will come in the next patch.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-14-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:03 +00:00
|
|
|
* Returns
|
|
|
|
|
* A pointer to an object of the type corresponding to the passed in
|
|
|
|
|
* 'local_type_id', or NULL on failure.
|
|
|
|
|
*/
|
|
|
|
|
extern void *bpf_obj_new_impl(__u64 local_type_id, void *meta) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_obj_new_impl */
|
|
|
|
|
#define bpf_obj_new(type) ((type *)bpf_obj_new_impl(bpf_core_type_id_local(type), NULL))
|
|
|
|
|
|
2022-11-18 01:56:04 +00:00
|
|
|
/* Description
|
|
|
|
|
* Free an allocated object. All fields of the object that require
|
|
|
|
|
* destruction will be destructed before the storage is freed.
|
|
|
|
|
*
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
* The 'meta' parameter is rewritten by the verifier, no need for BPF
|
|
|
|
|
* program to set it.
|
2022-11-18 01:56:04 +00:00
|
|
|
* Returns
|
|
|
|
|
* Void.
|
|
|
|
|
*/
|
|
|
|
|
extern void bpf_obj_drop_impl(void *kptr, void *meta) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_obj_drop_impl */
|
|
|
|
|
#define bpf_obj_drop(kptr) bpf_obj_drop_impl(kptr, NULL)
|
|
|
|
|
|
bpf: Add bpf_refcount_acquire kfunc
Currently, BPF programs can interact with the lifetime of refcounted
local kptrs in the following ways:
bpf_obj_new - Initialize refcount to 1 as part of new object creation
bpf_obj_drop - Decrement refcount and free object if it's 0
collection add - Pass ownership to the collection. No change to
refcount but collection is responsible for
bpf_obj_dropping it
In order to be able to add a refcounted local kptr to multiple
collections we need to be able to increment the refcount and acquire a
new owning reference. This patch adds a kfunc, bpf_refcount_acquire,
implementing such an operation.
bpf_refcount_acquire takes a refcounted local kptr and returns a new
owning reference to the same underlying memory as the input. The input
can be either owning or non-owning. To reinforce why this is safe,
consider the following code snippets:
struct node *n = bpf_obj_new(typeof(*n)); // A
struct node *m = bpf_refcount_acquire(n); // B
In the above snippet, n will be alive with refcount=1 after (A), and
since nothing changes that state before (B), it's obviously safe. If
n is instead added to some rbtree, we can still safely refcount_acquire
it:
struct node *n = bpf_obj_new(typeof(*n));
struct node *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&groot, &n->node, less); // A
m = bpf_refcount_acquire(n); // B
bpf_spin_unlock(&glock);
In the above snippet, after (A) n is a non-owning reference, and after
(B) m is an owning reference pointing to the same memory as n. Although
n has no ownership of that memory's lifetime, it's guaranteed to be
alive until the end of the critical section, and n would be clobbered if
we were past the end of the critical section, so it's safe to bump
refcount.
Implementation details:
* From verifier's perspective, bpf_refcount_acquire handling is similar
to bpf_obj_new and bpf_obj_drop. Like the former, it returns a new
owning reference matching input type, although like the latter, type
can be inferred from concrete kptr input. Verifier changes in
{check,fixup}_kfunc_call and check_kfunc_args are largely copied from
aforementioned functions' verifier changes.
* An exception to the above is the new KF_ARG_PTR_TO_REFCOUNTED_KPTR
arg, indicated by new "__refcounted_kptr" kfunc arg suffix. This is
necessary in order to handle both owning and non-owning input without
adding special-casing to "__alloc" arg handling. Also a convenient
place to confirm that input type has bpf_refcount field.
* The implemented kfunc is actually bpf_refcount_acquire_impl, with
'hidden' second arg that the verifier sets to the type's struct_meta
in fixup_kfunc_call.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-5-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:06 +00:00
|
|
|
/* Description
|
|
|
|
|
* Increment the refcount on a refcounted local kptr, turning the
|
|
|
|
|
* non-owning reference input into an owning reference in the process.
|
|
|
|
|
*
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
* The 'meta' parameter is rewritten by the verifier, no need for BPF
|
|
|
|
|
* program to set it.
|
bpf: Add bpf_refcount_acquire kfunc
Currently, BPF programs can interact with the lifetime of refcounted
local kptrs in the following ways:
bpf_obj_new - Initialize refcount to 1 as part of new object creation
bpf_obj_drop - Decrement refcount and free object if it's 0
collection add - Pass ownership to the collection. No change to
refcount but collection is responsible for
bpf_obj_dropping it
In order to be able to add a refcounted local kptr to multiple
collections we need to be able to increment the refcount and acquire a
new owning reference. This patch adds a kfunc, bpf_refcount_acquire,
implementing such an operation.
bpf_refcount_acquire takes a refcounted local kptr and returns a new
owning reference to the same underlying memory as the input. The input
can be either owning or non-owning. To reinforce why this is safe,
consider the following code snippets:
struct node *n = bpf_obj_new(typeof(*n)); // A
struct node *m = bpf_refcount_acquire(n); // B
In the above snippet, n will be alive with refcount=1 after (A), and
since nothing changes that state before (B), it's obviously safe. If
n is instead added to some rbtree, we can still safely refcount_acquire
it:
struct node *n = bpf_obj_new(typeof(*n));
struct node *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&groot, &n->node, less); // A
m = bpf_refcount_acquire(n); // B
bpf_spin_unlock(&glock);
In the above snippet, after (A) n is a non-owning reference, and after
(B) m is an owning reference pointing to the same memory as n. Although
n has no ownership of that memory's lifetime, it's guaranteed to be
alive until the end of the critical section, and n would be clobbered if
we were past the end of the critical section, so it's safe to bump
refcount.
Implementation details:
* From verifier's perspective, bpf_refcount_acquire handling is similar
to bpf_obj_new and bpf_obj_drop. Like the former, it returns a new
owning reference matching input type, although like the latter, type
can be inferred from concrete kptr input. Verifier changes in
{check,fixup}_kfunc_call and check_kfunc_args are largely copied from
aforementioned functions' verifier changes.
* An exception to the above is the new KF_ARG_PTR_TO_REFCOUNTED_KPTR
arg, indicated by new "__refcounted_kptr" kfunc arg suffix. This is
necessary in order to handle both owning and non-owning input without
adding special-casing to "__alloc" arg handling. Also a convenient
place to confirm that input type has bpf_refcount field.
* The implemented kfunc is actually bpf_refcount_acquire_impl, with
'hidden' second arg that the verifier sets to the type's struct_meta
in fixup_kfunc_call.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-5-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:06 +00:00
|
|
|
* Returns
|
|
|
|
|
* An owning reference to the object pointed to by 'kptr'
|
|
|
|
|
*/
|
|
|
|
|
extern void *bpf_refcount_acquire_impl(void *kptr, void *meta) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_refcount_acquire_impl */
|
|
|
|
|
#define bpf_refcount_acquire(kptr) bpf_refcount_acquire_impl(kptr, NULL)
|
|
|
|
|
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
/* Description
|
|
|
|
|
* Add a new entry to the beginning of the BPF linked list.
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
*
|
|
|
|
|
* The 'meta' and 'off' parameters are rewritten by the verifier, no need
|
|
|
|
|
* for BPF programs to set them
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
* Returns
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
* 0 if the node was successfully added
|
|
|
|
|
* -EINVAL if the node wasn't added because it's already in a list
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
*/
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
extern int bpf_list_push_front_impl(struct bpf_list_head *head,
|
|
|
|
|
struct bpf_list_node *node,
|
|
|
|
|
void *meta, __u64 off) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_list_push_front_impl */
|
|
|
|
|
#define bpf_list_push_front(head, node) bpf_list_push_front_impl(head, node, NULL, 0)
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Add a new entry to the end of the BPF linked list.
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
*
|
|
|
|
|
* The 'meta' and 'off' parameters are rewritten by the verifier, no need
|
|
|
|
|
* for BPF programs to set them
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
* Returns
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
* 0 if the node was successfully added
|
|
|
|
|
* -EINVAL if the node wasn't added because it's already in a list
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
*/
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
extern int bpf_list_push_back_impl(struct bpf_list_head *head,
|
|
|
|
|
struct bpf_list_node *node,
|
|
|
|
|
void *meta, __u64 off) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_list_push_back_impl */
|
|
|
|
|
#define bpf_list_push_back(head, node) bpf_list_push_back_impl(head, node, NULL, 0)
|
bpf: Introduce single ownership BPF linked list API
Add a linked list API for use in BPF programs, where it expects
protection from the bpf_spin_lock in the same allocation as the
bpf_list_head. For now, only one bpf_spin_lock can be present hence that
is assumed to be the one protecting the bpf_list_head.
The following functions are added to kick things off:
// Add node to beginning of list
void bpf_list_push_front(struct bpf_list_head *head, struct bpf_list_node *node);
// Add node to end of list
void bpf_list_push_back(struct bpf_list_head *head, struct bpf_list_node *node);
// Remove node at beginning of list and return it
struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head);
// Remove node at end of list and return it
struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head);
The lock protecting the bpf_list_head needs to be taken for all
operations. The verifier ensures that the lock that needs to be taken is
always held, and only the correct lock is taken for these operations.
These checks are made statically by relying on the reg->id preserved for
registers pointing into regions having both bpf_spin_lock and the
objects protected by it. The comment over check_reg_allocation_locked in
this change describes the logic in detail.
Note that bpf_list_push_front and bpf_list_push_back are meant to
consume the object containing the node in the 1st argument, however that
specific mechanism is intended to not release the ref_obj_id directly
until the bpf_spin_unlock is called. In this commit, nothing is done,
but the next commit will be introducing logic to handle this case, so it
has been left as is for now.
bpf_list_pop_front and bpf_list_pop_back delete the first or last item
of the list respectively, and return pointer to the element at the
list_node offset. The user can then use container_of style macro to get
the actual entry type. The verifier however statically knows the actual
type, so the safety properties are still preserved.
With these additions, programs can now manage their own linked lists and
store their objects in them.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-17-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:06 +00:00
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Remove the entry at the beginning of the BPF linked list.
|
|
|
|
|
* Returns
|
|
|
|
|
* Pointer to bpf_list_node of deleted entry, or NULL if list is empty.
|
|
|
|
|
*/
|
|
|
|
|
extern struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Remove the entry at the end of the BPF linked list.
|
|
|
|
|
* Returns
|
|
|
|
|
* Pointer to bpf_list_node of deleted entry, or NULL if list is empty.
|
|
|
|
|
*/
|
|
|
|
|
extern struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head) __ksym;
|
|
|
|
|
|
2023-02-14 00:40:15 +00:00
|
|
|
/* Description
|
|
|
|
|
* Remove 'node' from rbtree with root 'root'
|
|
|
|
|
* Returns
|
|
|
|
|
* Pointer to the removed node, or NULL if 'root' didn't contain 'node'
|
|
|
|
|
*/
|
|
|
|
|
extern struct bpf_rb_node *bpf_rbtree_remove(struct bpf_rb_root *root,
|
|
|
|
|
struct bpf_rb_node *node) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Add 'node' to rbtree with root 'root' using comparator 'less'
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
*
|
|
|
|
|
* The 'meta' and 'off' parameters are rewritten by the verifier, no need
|
|
|
|
|
* for BPF programs to set them
|
2023-02-14 00:40:15 +00:00
|
|
|
* Returns
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
* 0 if the node was successfully added
|
|
|
|
|
* -EINVAL if the node wasn't added because it's already in a tree
|
2023-02-14 00:40:15 +00:00
|
|
|
*/
|
bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Consider this code snippet:
struct node {
long key;
bpf_list_node l;
bpf_rb_node r;
bpf_refcount ref;
}
int some_bpf_prog(void *ctx)
{
struct node *n = bpf_obj_new(/*...*/), *m;
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->r, /* ... */);
m = bpf_refcount_acquire(n);
bpf_rbtree_add(&other_tree, &m->r, /* ... */);
bpf_spin_unlock(&glock);
/* ... */
}
After bpf_refcount_acquire, n and m point to the same underlying memory,
and that node's bpf_rb_node field is being used by the some_tree insert,
so overwriting it as a result of the second insert is an error. In order
to properly support refcounted nodes, the rbtree and list insert
functions must be allowed to fail. This patch adds such support.
The kfuncs bpf_rbtree_add, bpf_list_push_{front,back} are modified to
return an int indicating success/failure, with 0 -> success, nonzero ->
failure.
bpf_obj_drop on failure
=======================
Currently the only reason an insert can fail is the example above: the
bpf_{list,rb}_node is already in use. When such a failure occurs, the
insert kfuncs will bpf_obj_drop the input node. This allows the insert
operations to logically fail without changing their verifier owning ref
behavior, namely the unconditional release_reference of the input
owning ref.
With insert that always succeeds, ownership of the node is always passed
to the collection, since the node always ends up in the collection.
With a possibly-failed insert w/ bpf_obj_drop, ownership of the node
is always passed either to the collection (success), or to bpf_obj_drop
(failure). Regardless, it's correct to continue unconditionally
releasing the input owning ref, as something is always taking ownership
from the calling program on insert.
Keeping owning ref behavior unchanged results in a nice default UX for
insert functions that can fail. If the program's reaction to a failed
insert is "fine, just get rid of this owning ref for me and let me go
on with my business", then there's no reason to check for failure since
that's default behavior. e.g.:
long important_failures = 0;
int some_bpf_prog(void *ctx)
{
struct node *n, *m, *o; /* all bpf_obj_new'd */
bpf_spin_lock(&glock);
bpf_rbtree_add(&some_tree, &n->node, /* ... */);
bpf_rbtree_add(&some_tree, &m->node, /* ... */);
if (bpf_rbtree_add(&some_tree, &o->node, /* ... */)) {
important_failures++;
}
bpf_spin_unlock(&glock);
}
If we instead chose to pass ownership back to the program on failed
insert - by returning NULL on success or an owning ref on failure -
programs would always have to do something with the returned ref on
failure. The most likely action is probably "I'll just get rid of this
owning ref and go about my business", which ideally would look like:
if (n = bpf_rbtree_add(&some_tree, &n->node, /* ... */))
bpf_obj_drop(n);
But bpf_obj_drop isn't allowed in a critical section and inserts must
occur within one, so in reality error handling would become a
hard-to-parse mess.
For refcounted nodes, we can replicate the "pass ownership back to
program on failure" logic with this patch's semantics, albeit in an ugly
way:
struct node *n = bpf_obj_new(/* ... */), *m;
bpf_spin_lock(&glock);
m = bpf_refcount_acquire(n);
if (bpf_rbtree_add(&some_tree, &n->node, /* ... */)) {
/* Do something with m */
}
bpf_spin_unlock(&glock);
bpf_obj_drop(m);
bpf_refcount_acquire is used to simulate "return owning ref on failure".
This should be an uncommon occurrence, though.
Addition of two verifier-fixup'd args to collection inserts
===========================================================
The actual bpf_obj_drop kfunc is
bpf_obj_drop_impl(void *, struct btf_struct_meta *), with bpf_obj_drop
macro populating the second arg with 0 and the verifier later filling in
the arg during insn fixup.
Because bpf_rbtree_add and bpf_list_push_{front,back} now might do
bpf_obj_drop, these kfuncs need a btf_struct_meta parameter that can be
passed to bpf_obj_drop_impl.
Similarly, because the 'node' param to those insert functions is the
bpf_{list,rb}_node within the node type, and bpf_obj_drop expects a
pointer to the beginning of the node, the insert functions need to be
able to find the beginning of the node struct. A second
verifier-populated param is necessary: the offset of {list,rb}_node within the
node type.
These two new params allow the insert kfuncs to correctly call
__bpf_obj_drop_impl:
beginning_of_node = bpf_rb_node_ptr - offset
if (already_inserted)
__bpf_obj_drop_impl(beginning_of_node, btf_struct_meta->record);
Similarly to other kfuncs with "hidden" verifier-populated params, the
insert functions are renamed with _impl prefix and a macro is provided
for common usage. For example, bpf_rbtree_add kfunc is now
bpf_rbtree_add_impl and bpf_rbtree_add is now a macro which sets
"hidden" args to 0.
Due to the two new args BPF progs will need to be recompiled to work
with the new _impl kfuncs.
This patch also rewrites the "hidden argument" explanation to more
directly say why the BPF program writer doesn't need to populate the
arguments with anything meaningful.
How does this new logic affect non-owning references?
=====================================================
Currently, non-owning refs are valid until the end of the critical
section in which they're created. We can make this guarantee because, if
a non-owning ref exists, the referent was added to some collection. The
collection will drop() its nodes when it goes away, but it can't go away
while our program is accessing it, so that's not a problem. If the
referent is removed from the collection in the same CS that it was added
in, it can't be bpf_obj_drop'd until after CS end. Those are the only
two ways to free the referent's memory and neither can happen until
after the non-owning ref's lifetime ends.
On first glance, having these collection insert functions potentially
bpf_obj_drop their input seems like it breaks the "can't be
bpf_obj_drop'd until after CS end" line of reasoning. But we care about
the memory not being _freed_ until end of CS end, and a previous patch
in the series modified bpf_obj_drop such that it doesn't free refcounted
nodes until refcount == 0. So the statement can be more accurately
rewritten as "can't be free'd until after CS end".
We can prove that this rewritten statement holds for any non-owning
reference produced by collection insert functions:
* If the input to the insert function is _not_ refcounted
* We have an owning reference to the input, and can conclude it isn't
in any collection
* Inserting a node in a collection turns owning refs into
non-owning, and since our input type isn't refcounted, there's no
way to obtain additional owning refs to the same underlying
memory
* Because our node isn't in any collection, the insert operation
cannot fail, so bpf_obj_drop will not execute
* If bpf_obj_drop is guaranteed not to execute, there's no risk of
memory being free'd
* Otherwise, the input to the insert function is refcounted
* If the insert operation fails due to the node's list_head or rb_root
already being in some collection, there was some previous successful
insert which passed refcount to the collection
* We have an owning reference to the input, it must have been
acquired via bpf_refcount_acquire, which bumped the refcount
* refcount must be >= 2 since there's a valid owning reference and the
node is already in a collection
* Insert triggering bpf_obj_drop will decr refcount to >= 1, never
resulting in a free
So although we may do bpf_obj_drop during the critical section, this
will never result in memory being free'd, and no changes to non-owning
ref logic are needed in this patch.
Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20230415201811.343116-6-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-04-15 20:18:07 +00:00
|
|
|
extern int bpf_rbtree_add_impl(struct bpf_rb_root *root, struct bpf_rb_node *node,
|
|
|
|
|
bool (less)(struct bpf_rb_node *a, const struct bpf_rb_node *b),
|
|
|
|
|
void *meta, __u64 off) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_rbtree_add_impl */
|
|
|
|
|
#define bpf_rbtree_add(head, node, less) bpf_rbtree_add_impl(head, node, less, NULL, 0)
|
2023-02-14 00:40:15 +00:00
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Return the first (leftmost) node in input tree
|
|
|
|
|
* Returns
|
|
|
|
|
* Pointer to the node, which is _not_ removed from the tree. If the tree
|
|
|
|
|
* contains no nodes, returns NULL.
|
|
|
|
|
*/
|
|
|
|
|
extern struct bpf_rb_node *bpf_rbtree_first(struct bpf_rb_root *root) __ksym;
|
|
|
|
|
|
2023-08-27 15:28:05 +00:00
|
|
|
/* Description
|
|
|
|
|
* Allocates a percpu object of the type represented by 'local_type_id' in
|
|
|
|
|
* program BTF. User may use the bpf_core_type_id_local macro to pass the
|
|
|
|
|
* type ID of a struct in program BTF.
|
|
|
|
|
*
|
|
|
|
|
* The 'local_type_id' parameter must be a known constant.
|
|
|
|
|
* The 'meta' parameter is rewritten by the verifier, no need for BPF
|
|
|
|
|
* program to set it.
|
|
|
|
|
* Returns
|
|
|
|
|
* A pointer to a percpu object of the type corresponding to the passed in
|
|
|
|
|
* 'local_type_id', or NULL on failure.
|
|
|
|
|
*/
|
|
|
|
|
extern void *bpf_percpu_obj_new_impl(__u64 local_type_id, void *meta) __ksym;
|
|
|
|
|
|
|
|
|
|
/* Convenience macro to wrap over bpf_percpu_obj_new_impl */
|
|
|
|
|
#define bpf_percpu_obj_new(type) ((type __percpu_kptr *)bpf_percpu_obj_new_impl(bpf_core_type_id_local(type), NULL))
|
|
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Free an allocated percpu object. All fields of the object that require
|
|
|
|
|
* destruction will be destructed before the storage is freed.
|
|
|
|
|
*
|
|
|
|
|
* The 'meta' parameter is rewritten by the verifier, no need for BPF
|
|
|
|
|
* program to set it.
|
|
|
|
|
* Returns
|
|
|
|
|
* Void.
|
|
|
|
|
*/
|
|
|
|
|
extern void bpf_percpu_obj_drop_impl(void *kptr, void *meta) __ksym;
|
|
|
|
|
|
2023-10-13 20:44:25 +00:00
|
|
|
struct bpf_iter_task_vma;
|
|
|
|
|
|
|
|
|
|
extern int bpf_iter_task_vma_new(struct bpf_iter_task_vma *it,
|
|
|
|
|
struct task_struct *task,
|
|
|
|
|
unsigned long addr) __ksym;
|
|
|
|
|
extern struct vm_area_struct *bpf_iter_task_vma_next(struct bpf_iter_task_vma *it) __ksym;
|
|
|
|
|
extern void bpf_iter_task_vma_destroy(struct bpf_iter_task_vma *it) __ksym;
|
|
|
|
|
|
2023-08-27 15:28:05 +00:00
|
|
|
/* Convenience macro to wrap over bpf_obj_drop_impl */
|
|
|
|
|
#define bpf_percpu_obj_drop(kptr) bpf_percpu_obj_drop_impl(kptr, NULL)
|
|
|
|
|
|
bpf: Implement BPF exceptions
This patch implements BPF exceptions, and introduces a bpf_throw kfunc
to allow programs to throw exceptions during their execution at runtime.
A bpf_throw invocation is treated as an immediate termination of the
program, returning back to its caller within the kernel, unwinding all
stack frames.
This allows the program to simplify its implementation, by testing for
runtime conditions which the verifier has no visibility into, and assert
that they are true. In case they are not, the program can simply throw
an exception from the other branch.
BPF exceptions are explicitly *NOT* an unlikely slowpath error handling
primitive, and this objective has guided design choices of the
implementation of the them within the kernel (with the bulk of the cost
for unwinding the stack offloaded to the bpf_throw kfunc).
The implementation of this mechanism requires use of add_hidden_subprog
mechanism introduced in the previous patch, which generates a couple of
instructions to move R1 to R0 and exit. The JIT then rewrites the
prologue of this subprog to take the stack pointer and frame pointer as
inputs and reset the stack frame, popping all callee-saved registers
saved by the main subprog. The bpf_throw function then walks the stack
at runtime, and invokes this exception subprog with the stack and frame
pointers as parameters.
Reviewers must take note that currently the main program is made to save
all callee-saved registers on x86_64 during entry into the program. This
is because we must do an equivalent of a lightweight context switch when
unwinding the stack, therefore we need the callee-saved registers of the
caller of the BPF program to be able to return with a sane state.
Note that we have to additionally handle r12, even though it is not used
by the program, because when throwing the exception the program makes an
entry into the kernel which could clobber r12 after saving it on the
stack. To be able to preserve the value we received on program entry, we
push r12 and restore it from the generated subprogram when unwinding the
stack.
For now, bpf_throw invocation fails when lingering resources or locks
exist in that path of the program. In a future followup, bpf_throw will
be extended to perform frame-by-frame unwinding to release lingering
resources for each stack frame, removing this limitation.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-5-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-12 23:32:01 +00:00
|
|
|
/* Description
|
|
|
|
|
* Throw a BPF exception from the program, immediately terminating its
|
|
|
|
|
* execution and unwinding the stack. The supplied 'cookie' parameter
|
bpf: Add support for custom exception callbacks
By default, the subprog generated by the verifier to handle a thrown
exception hardcodes a return value of 0. To allow user-defined logic
and modification of the return value when an exception is thrown,
introduce the 'exception_callback:' declaration tag, which marks a
callback as the default exception handler for the program.
The format of the declaration tag is 'exception_callback:<value>', where
<value> is the name of the exception callback. Each main program can be
tagged using this BTF declaratiion tag to associate it with an exception
callback. In case the tag is absent, the default callback is used.
As such, the exception callback cannot be modified at runtime, only set
during verification.
Allowing modification of the callback for the current program execution
at runtime leads to issues when the programs begin to nest, as any
per-CPU state maintaing this information will have to be saved and
restored. We don't want it to stay in bpf_prog_aux as this takes a
global effect for all programs. An alternative solution is spilling
the callback pointer at a known location on the program stack on entry,
and then passing this location to bpf_throw as a parameter.
However, since exceptions are geared more towards a use case where they
are ideally never invoked, optimizing for this use case and adding to
the complexity has diminishing returns.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-7-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-12 23:32:03 +00:00
|
|
|
* will be the return value of the program when an exception is thrown,
|
|
|
|
|
* and the default exception callback is used. Otherwise, if an exception
|
|
|
|
|
* callback is set using the '__exception_cb(callback)' declaration tag
|
|
|
|
|
* on the main program, the 'cookie' parameter will be the callback's only
|
|
|
|
|
* input argument.
|
|
|
|
|
*
|
|
|
|
|
* Thus, in case of default exception callback, 'cookie' is subjected to
|
|
|
|
|
* constraints on the program's return value (as with R0 on exit).
|
|
|
|
|
* Otherwise, the return value of the marked exception callback will be
|
|
|
|
|
* subjected to the same checks.
|
bpf: Implement BPF exceptions
This patch implements BPF exceptions, and introduces a bpf_throw kfunc
to allow programs to throw exceptions during their execution at runtime.
A bpf_throw invocation is treated as an immediate termination of the
program, returning back to its caller within the kernel, unwinding all
stack frames.
This allows the program to simplify its implementation, by testing for
runtime conditions which the verifier has no visibility into, and assert
that they are true. In case they are not, the program can simply throw
an exception from the other branch.
BPF exceptions are explicitly *NOT* an unlikely slowpath error handling
primitive, and this objective has guided design choices of the
implementation of the them within the kernel (with the bulk of the cost
for unwinding the stack offloaded to the bpf_throw kfunc).
The implementation of this mechanism requires use of add_hidden_subprog
mechanism introduced in the previous patch, which generates a couple of
instructions to move R1 to R0 and exit. The JIT then rewrites the
prologue of this subprog to take the stack pointer and frame pointer as
inputs and reset the stack frame, popping all callee-saved registers
saved by the main subprog. The bpf_throw function then walks the stack
at runtime, and invokes this exception subprog with the stack and frame
pointers as parameters.
Reviewers must take note that currently the main program is made to save
all callee-saved registers on x86_64 during entry into the program. This
is because we must do an equivalent of a lightweight context switch when
unwinding the stack, therefore we need the callee-saved registers of the
caller of the BPF program to be able to return with a sane state.
Note that we have to additionally handle r12, even though it is not used
by the program, because when throwing the exception the program makes an
entry into the kernel which could clobber r12 after saving it on the
stack. To be able to preserve the value we received on program entry, we
push r12 and restore it from the generated subprogram when unwinding the
stack.
For now, bpf_throw invocation fails when lingering resources or locks
exist in that path of the program. In a future followup, bpf_throw will
be extended to perform frame-by-frame unwinding to release lingering
resources for each stack frame, removing this limitation.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-5-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-12 23:32:01 +00:00
|
|
|
*
|
|
|
|
|
* Note that throwing an exception with lingering resources (locks,
|
|
|
|
|
* references, etc.) will lead to a verification error.
|
|
|
|
|
*
|
|
|
|
|
* Note that callbacks *cannot* call this helper.
|
|
|
|
|
* Returns
|
|
|
|
|
* Never.
|
|
|
|
|
* Throws
|
|
|
|
|
* An exception with the specified 'cookie' value.
|
|
|
|
|
*/
|
|
|
|
|
extern void bpf_throw(u64 cookie) __ksym;
|
|
|
|
|
|
bpf: Add support for custom exception callbacks
By default, the subprog generated by the verifier to handle a thrown
exception hardcodes a return value of 0. To allow user-defined logic
and modification of the return value when an exception is thrown,
introduce the 'exception_callback:' declaration tag, which marks a
callback as the default exception handler for the program.
The format of the declaration tag is 'exception_callback:<value>', where
<value> is the name of the exception callback. Each main program can be
tagged using this BTF declaratiion tag to associate it with an exception
callback. In case the tag is absent, the default callback is used.
As such, the exception callback cannot be modified at runtime, only set
during verification.
Allowing modification of the callback for the current program execution
at runtime leads to issues when the programs begin to nest, as any
per-CPU state maintaing this information will have to be saved and
restored. We don't want it to stay in bpf_prog_aux as this takes a
global effect for all programs. An alternative solution is spilling
the callback pointer at a known location on the program stack on entry,
and then passing this location to bpf_throw as a parameter.
However, since exceptions are geared more towards a use case where they
are ideally never invoked, optimizing for this use case and adding to
the complexity has diminishing returns.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20230912233214.1518551-7-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-09-12 23:32:03 +00:00
|
|
|
/* This macro must be used to mark the exception callback corresponding to the
|
|
|
|
|
* main program. For example:
|
|
|
|
|
*
|
|
|
|
|
* int exception_cb(u64 cookie) {
|
|
|
|
|
* return cookie;
|
|
|
|
|
* }
|
|
|
|
|
*
|
|
|
|
|
* SEC("tc")
|
|
|
|
|
* __exception_cb(exception_cb)
|
|
|
|
|
* int main_prog(struct __sk_buff *ctx) {
|
|
|
|
|
* ...
|
|
|
|
|
* return TC_ACT_OK;
|
|
|
|
|
* }
|
|
|
|
|
*
|
|
|
|
|
* Here, exception callback for the main program will be 'exception_cb'. Note
|
|
|
|
|
* that this attribute can only be used once, and multiple exception callbacks
|
|
|
|
|
* specified for the main program will lead to verification error.
|
|
|
|
|
*/
|
|
|
|
|
#define __exception_cb(name) __attribute__((btf_decl_tag("exception_callback:" #name)))
|
|
|
|
|
|
2023-09-12 23:32:13 +00:00
|
|
|
#define __bpf_assert_signed(x) _Generic((x), \
|
|
|
|
|
unsigned long: 0, \
|
|
|
|
|
unsigned long long: 0, \
|
|
|
|
|
signed long: 1, \
|
|
|
|
|
signed long long: 1 \
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
#define __bpf_assert_check(LHS, op, RHS) \
|
|
|
|
|
_Static_assert(sizeof(&(LHS)), "1st argument must be an lvalue expression"); \
|
|
|
|
|
_Static_assert(sizeof(LHS) == 8, "Only 8-byte integers are supported\n"); \
|
|
|
|
|
_Static_assert(__builtin_constant_p(__bpf_assert_signed(LHS)), "internal static assert"); \
|
|
|
|
|
_Static_assert(__builtin_constant_p((RHS)), "2nd argument must be a constant expression")
|
|
|
|
|
|
|
|
|
|
#define __bpf_assert(LHS, op, cons, RHS, VAL) \
|
|
|
|
|
({ \
|
|
|
|
|
(void)bpf_throw; \
|
|
|
|
|
asm volatile ("if %[lhs] " op " %[rhs] goto +2; r1 = %[value]; call bpf_throw" \
|
|
|
|
|
: : [lhs] "r"(LHS), [rhs] cons(RHS), [value] "ri"(VAL) : ); \
|
|
|
|
|
})
|
|
|
|
|
|
|
|
|
|
#define __bpf_assert_op_sign(LHS, op, cons, RHS, VAL, supp_sign) \
|
|
|
|
|
({ \
|
|
|
|
|
__bpf_assert_check(LHS, op, RHS); \
|
|
|
|
|
if (__bpf_assert_signed(LHS) && !(supp_sign)) \
|
|
|
|
|
__bpf_assert(LHS, "s" #op, cons, RHS, VAL); \
|
|
|
|
|
else \
|
|
|
|
|
__bpf_assert(LHS, #op, cons, RHS, VAL); \
|
|
|
|
|
})
|
|
|
|
|
|
|
|
|
|
#define __bpf_assert_op(LHS, op, RHS, VAL, supp_sign) \
|
|
|
|
|
({ \
|
|
|
|
|
if (sizeof(typeof(RHS)) == 8) { \
|
|
|
|
|
const typeof(RHS) rhs_var = (RHS); \
|
|
|
|
|
__bpf_assert_op_sign(LHS, op, "r", rhs_var, VAL, supp_sign); \
|
|
|
|
|
} else { \
|
|
|
|
|
__bpf_assert_op_sign(LHS, op, "i", RHS, VAL, supp_sign); \
|
|
|
|
|
} \
|
|
|
|
|
})
|
|
|
|
|
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
#define __cmp_cannot_be_signed(x) \
|
|
|
|
|
__builtin_strcmp(#x, "==") == 0 || __builtin_strcmp(#x, "!=") == 0 || \
|
|
|
|
|
__builtin_strcmp(#x, "&") == 0
|
|
|
|
|
|
|
|
|
|
#define __is_signed_type(type) (((type)(-1)) < (type)1)
|
|
|
|
|
|
2024-01-12 22:01:34 +00:00
|
|
|
#define __bpf_cmp(LHS, OP, PRED, RHS, DEFAULT) \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
({ \
|
|
|
|
|
__label__ l_true; \
|
|
|
|
|
bool ret = DEFAULT; \
|
2024-01-12 22:01:34 +00:00
|
|
|
asm volatile goto("if %[lhs] " OP " %[rhs] goto %l[l_true]" \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
:: [lhs] "r"((short)LHS), [rhs] PRED (RHS) :: l_true); \
|
|
|
|
|
ret = !DEFAULT; \
|
|
|
|
|
l_true: \
|
|
|
|
|
ret; \
|
|
|
|
|
})
|
|
|
|
|
|
|
|
|
|
/* C type conversions coupled with comparison operator are tricky.
|
|
|
|
|
* Make sure BPF program is compiled with -Wsign-compare then
|
|
|
|
|
* __lhs OP __rhs below will catch the mistake.
|
|
|
|
|
* Be aware that we check only __lhs to figure out the sign of compare.
|
|
|
|
|
*/
|
2024-01-12 22:01:34 +00:00
|
|
|
#define _bpf_cmp(LHS, OP, RHS, UNLIKELY) \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
({ \
|
|
|
|
|
typeof(LHS) __lhs = (LHS); \
|
|
|
|
|
typeof(RHS) __rhs = (RHS); \
|
|
|
|
|
bool ret; \
|
|
|
|
|
_Static_assert(sizeof(&(LHS)), "1st argument must be an lvalue expression"); \
|
|
|
|
|
(void)(__lhs OP __rhs); \
|
|
|
|
|
if (__cmp_cannot_be_signed(OP) || !__is_signed_type(typeof(__lhs))) { \
|
|
|
|
|
if (sizeof(__rhs) == 8) \
|
2024-01-12 22:01:34 +00:00
|
|
|
/* "i" will truncate 64-bit constant into s32, \
|
|
|
|
|
* so we have to use extra register via "r". \
|
|
|
|
|
*/ \
|
|
|
|
|
ret = __bpf_cmp(__lhs, #OP, "r", __rhs, UNLIKELY); \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
else \
|
2024-01-12 22:01:34 +00:00
|
|
|
ret = __bpf_cmp(__lhs, #OP, "ri", __rhs, UNLIKELY); \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
} else { \
|
|
|
|
|
if (sizeof(__rhs) == 8) \
|
2024-01-12 22:01:34 +00:00
|
|
|
ret = __bpf_cmp(__lhs, "s"#OP, "r", __rhs, UNLIKELY); \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
else \
|
2024-01-12 22:01:34 +00:00
|
|
|
ret = __bpf_cmp(__lhs, "s"#OP, "ri", __rhs, UNLIKELY); \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
} \
|
|
|
|
|
ret; \
|
|
|
|
|
})
|
|
|
|
|
|
|
|
|
|
#ifndef bpf_cmp_unlikely
|
|
|
|
|
#define bpf_cmp_unlikely(LHS, OP, RHS) _bpf_cmp(LHS, OP, RHS, true)
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
#ifndef bpf_cmp_likely
|
|
|
|
|
#define bpf_cmp_likely(LHS, OP, RHS) \
|
|
|
|
|
({ \
|
2024-01-12 22:01:34 +00:00
|
|
|
bool ret = 0; \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
if (__builtin_strcmp(#OP, "==") == 0) \
|
|
|
|
|
ret = _bpf_cmp(LHS, !=, RHS, false); \
|
|
|
|
|
else if (__builtin_strcmp(#OP, "!=") == 0) \
|
|
|
|
|
ret = _bpf_cmp(LHS, ==, RHS, false); \
|
|
|
|
|
else if (__builtin_strcmp(#OP, "<=") == 0) \
|
|
|
|
|
ret = _bpf_cmp(LHS, >, RHS, false); \
|
|
|
|
|
else if (__builtin_strcmp(#OP, "<") == 0) \
|
|
|
|
|
ret = _bpf_cmp(LHS, >=, RHS, false); \
|
|
|
|
|
else if (__builtin_strcmp(#OP, ">") == 0) \
|
|
|
|
|
ret = _bpf_cmp(LHS, <=, RHS, false); \
|
|
|
|
|
else if (__builtin_strcmp(#OP, ">=") == 0) \
|
|
|
|
|
ret = _bpf_cmp(LHS, <, RHS, false); \
|
|
|
|
|
else \
|
2024-01-12 22:01:34 +00:00
|
|
|
asm volatile("r0 " #OP " invalid compare"); \
|
bpf: Introduce "volatile compare" macros
Compilers optimize conditional operators at will, but often bpf programmers
want to force compilers to keep the same operator in asm as it's written in C.
Introduce bpf_cmp_likely/unlikely(var1, conditional_op, var2) macros that can be used as:
- if (seen >= 1000)
+ if (bpf_cmp_unlikely(seen, >=, 1000))
The macros take advantage of BPF assembly that is C like.
The macros check the sign of variable 'seen' and emits either
signed or unsigned compare.
For example:
int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX s> 0 goto' in BPF assembly.
unsigned int a;
bpf_cmp_unlikely(a, >, 0) will be translated to 'if rX > 0 goto' in BPF assembly.
C type conversions coupled with comparison operator are tricky.
int i = -1;
unsigned int j = 1;
if (i < j) // this is false.
long i = -1;
unsigned int j = 1;
if (i < j) // this is true.
Make sure BPF program is compiled with -Wsign-compare then the macros will catch
the mistake.
The macros check LHS (left hand side) only to figure out the sign of compare.
'if 0 < rX goto' is not allowed in the assembly, so the users
have to use a variable on LHS anyway.
The patch updates few tests to demonstrate the use of the macros.
The macro allows to use BPF_JSET in C code, since LLVM doesn't generate it at
present. For example:
if (i & j) compiles into r0 &= r1; if r0 == 0 goto
while
if (bpf_cmp_unlikely(i, &, j)) compiles into if r0 & r1 goto
Note that the macros has to be careful with RHS assembly predicate.
Since:
u64 __rhs = 1ull << 42;
asm goto("if r0 < %[rhs] goto +1" :: [rhs] "ri" (__rhs));
LLVM will silently truncate 64-bit constant into s32 imm.
Note that [lhs] "r"((short)LHS) the type cast is a workaround for LLVM issue.
When LHS is exactly 32-bit LLVM emits redundant <<=32, >>=32 to zero upper 32-bits.
When LHS is 64 or 16 or 8-bit variable there are no shifts.
When LHS is 32-bit the (u64) cast doesn't help. Hence use (short) cast.
It does _not_ truncate the variable before it's assigned to a register.
Traditional likely()/unlikely() macros that use __builtin_expect(!!(x), 1 or 0)
have no effect on these macros, hence macros implement the logic manually.
bpf_cmp_unlikely() macro preserves compare operator as-is while
bpf_cmp_likely() macro flips the compare.
Consider two cases:
A.
for() {
if (foo >= 10) {
bar += foo;
}
other code;
}
B.
for() {
if (foo >= 10)
break;
other code;
}
It's ok to use either bpf_cmp_likely or bpf_cmp_unlikely macros in both cases,
but consider that 'break' is effectively 'goto out_of_the_loop'.
Hence it's better to use bpf_cmp_unlikely in the B case.
While 'bar += foo' is better to keep as 'fallthrough' == likely code path in the A case.
When it's written as:
A.
for() {
if (bpf_cmp_likely(foo, >=, 10)) {
bar += foo;
}
other code;
}
B.
for() {
if (bpf_cmp_unlikely(foo, >=, 10))
break;
other code;
}
The assembly will look like:
A.
for() {
if r1 < 10 goto L1;
bar += foo;
L1:
other code;
}
B.
for() {
if r1 >= 10 goto L2;
other code;
}
L2:
The bpf_cmp_likely vs bpf_cmp_unlikely changes basic block layout, hence it will
greatly influence the verification process. The number of processed instructions
will be different, since the verifier walks the fallthrough first.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20231226191148.48536-3-alexei.starovoitov@gmail.com
2023-12-26 19:11:44 +00:00
|
|
|
ret; \
|
|
|
|
|
})
|
|
|
|
|
#endif
|
|
|
|
|
|
2024-03-06 03:19:28 +00:00
|
|
|
#define cond_break \
|
|
|
|
|
({ __label__ l_break, l_continue; \
|
|
|
|
|
asm volatile goto("1:.byte 0xe5; \
|
|
|
|
|
.byte 0; \
|
|
|
|
|
.long ((%l[l_break] - 1b - 8) / 8) & 0xffff; \
|
|
|
|
|
.short 0" \
|
|
|
|
|
:::: l_break); \
|
|
|
|
|
goto l_continue; \
|
|
|
|
|
l_break: break; \
|
|
|
|
|
l_continue:; \
|
|
|
|
|
})
|
|
|
|
|
|
2023-12-26 19:11:47 +00:00
|
|
|
#ifndef bpf_nop_mov
|
|
|
|
|
#define bpf_nop_mov(var) \
|
|
|
|
|
asm volatile("%[reg]=%[reg]"::[reg]"r"((short)var))
|
|
|
|
|
#endif
|
|
|
|
|
|
2024-03-08 01:08:09 +00:00
|
|
|
/* emit instruction:
|
|
|
|
|
* rX = rX .off = BPF_ADDR_SPACE_CAST .imm32 = (dst_as << 16) | src_as
|
|
|
|
|
*/
|
|
|
|
|
#ifndef bpf_addr_space_cast
|
|
|
|
|
#define bpf_addr_space_cast(var, dst_as, src_as)\
|
|
|
|
|
asm volatile(".byte 0xBF; \
|
|
|
|
|
.ifc %[reg], r0; \
|
|
|
|
|
.byte 0x00; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r1; \
|
|
|
|
|
.byte 0x11; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r2; \
|
|
|
|
|
.byte 0x22; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r3; \
|
|
|
|
|
.byte 0x33; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r4; \
|
|
|
|
|
.byte 0x44; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r5; \
|
|
|
|
|
.byte 0x55; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r6; \
|
|
|
|
|
.byte 0x66; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r7; \
|
|
|
|
|
.byte 0x77; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r8; \
|
|
|
|
|
.byte 0x88; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.ifc %[reg], r9; \
|
|
|
|
|
.byte 0x99; \
|
|
|
|
|
.endif; \
|
|
|
|
|
.short %[off]; \
|
|
|
|
|
.long %[as]" \
|
|
|
|
|
: [reg]"+r"(var) \
|
|
|
|
|
: [off]"i"(BPF_ADDR_SPACE_CAST) \
|
|
|
|
|
, [as]"i"((dst_as << 16) | src_as));
|
|
|
|
|
#endif
|
|
|
|
|
|
2023-09-12 23:32:13 +00:00
|
|
|
/* Description
|
|
|
|
|
* Assert that a conditional expression is true.
|
|
|
|
|
* Returns
|
|
|
|
|
* Void.
|
|
|
|
|
* Throws
|
|
|
|
|
* An exception with the value zero when the assertion fails.
|
|
|
|
|
*/
|
|
|
|
|
#define bpf_assert(cond) if (!(cond)) bpf_throw(0);
|
|
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Assert that a conditional expression is true.
|
|
|
|
|
* Returns
|
|
|
|
|
* Void.
|
|
|
|
|
* Throws
|
|
|
|
|
* An exception with the specified value when the assertion fails.
|
|
|
|
|
*/
|
|
|
|
|
#define bpf_assert_with(cond, value) if (!(cond)) bpf_throw(value);
|
|
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Assert that LHS is in the range [BEG, END] (inclusive of both). This
|
|
|
|
|
* statement updates the known bounds of LHS during verification. Note
|
|
|
|
|
* that both BEG and END must be constant values, and must fit within the
|
|
|
|
|
* data type of LHS.
|
|
|
|
|
* Returns
|
|
|
|
|
* Void.
|
|
|
|
|
* Throws
|
|
|
|
|
* An exception with the value zero when the assertion fails.
|
|
|
|
|
*/
|
|
|
|
|
#define bpf_assert_range(LHS, BEG, END) \
|
|
|
|
|
({ \
|
|
|
|
|
_Static_assert(BEG <= END, "BEG must be <= END"); \
|
|
|
|
|
barrier_var(LHS); \
|
|
|
|
|
__bpf_assert_op(LHS, >=, BEG, 0, false); \
|
|
|
|
|
__bpf_assert_op(LHS, <=, END, 0, false); \
|
|
|
|
|
})
|
|
|
|
|
|
|
|
|
|
/* Description
|
|
|
|
|
* Assert that LHS is in the range [BEG, END] (inclusive of both). This
|
|
|
|
|
* statement updates the known bounds of LHS during verification. Note
|
|
|
|
|
* that both BEG and END must be constant values, and must fit within the
|
|
|
|
|
* data type of LHS.
|
|
|
|
|
* Returns
|
|
|
|
|
* Void.
|
|
|
|
|
* Throws
|
|
|
|
|
* An exception with the specified value when the assertion fails.
|
|
|
|
|
*/
|
|
|
|
|
#define bpf_assert_range_with(LHS, BEG, END, value) \
|
|
|
|
|
({ \
|
|
|
|
|
_Static_assert(BEG <= END, "BEG must be <= END"); \
|
|
|
|
|
barrier_var(LHS); \
|
|
|
|
|
__bpf_assert_op(LHS, >=, BEG, value, false); \
|
|
|
|
|
__bpf_assert_op(LHS, <=, END, value, false); \
|
|
|
|
|
})
|
|
|
|
|
|
2023-10-18 06:17:40 +00:00
|
|
|
struct bpf_iter_css_task;
|
|
|
|
|
struct cgroup_subsys_state;
|
|
|
|
|
extern int bpf_iter_css_task_new(struct bpf_iter_css_task *it,
|
|
|
|
|
struct cgroup_subsys_state *css, unsigned int flags) __weak __ksym;
|
|
|
|
|
extern struct task_struct *bpf_iter_css_task_next(struct bpf_iter_css_task *it) __weak __ksym;
|
|
|
|
|
extern void bpf_iter_css_task_destroy(struct bpf_iter_css_task *it) __weak __ksym;
|
|
|
|
|
|
2023-10-18 06:17:41 +00:00
|
|
|
struct bpf_iter_task;
|
|
|
|
|
extern int bpf_iter_task_new(struct bpf_iter_task *it,
|
|
|
|
|
struct task_struct *task, unsigned int flags) __weak __ksym;
|
|
|
|
|
extern struct task_struct *bpf_iter_task_next(struct bpf_iter_task *it) __weak __ksym;
|
|
|
|
|
extern void bpf_iter_task_destroy(struct bpf_iter_task *it) __weak __ksym;
|
2023-10-18 06:17:40 +00:00
|
|
|
|
2023-10-18 06:17:42 +00:00
|
|
|
struct bpf_iter_css;
|
|
|
|
|
extern int bpf_iter_css_new(struct bpf_iter_css *it,
|
|
|
|
|
struct cgroup_subsys_state *start, unsigned int flags) __weak __ksym;
|
|
|
|
|
extern struct cgroup_subsys_state *bpf_iter_css_next(struct bpf_iter_css *it) __weak __ksym;
|
|
|
|
|
extern void bpf_iter_css_destroy(struct bpf_iter_css *it) __weak __ksym;
|
|
|
|
|
|
bpf: Introduce bpf_obj_new
Introduce type safe memory allocator bpf_obj_new for BPF programs. The
kernel side kfunc is named bpf_obj_new_impl, as passing hidden arguments
to kfuncs still requires having them in prototype, unlike BPF helpers
which always take 5 arguments and have them checked using bpf_func_proto
in verifier, ignoring unset argument types.
Introduce __ign suffix to ignore a specific kfunc argument during type
checks, then use this to introduce support for passing type metadata to
the bpf_obj_new_impl kfunc.
The user passes BTF ID of the type it wants to allocates in program BTF,
the verifier then rewrites the first argument as the size of this type,
after performing some sanity checks (to ensure it exists and it is a
struct type).
The second argument is also fixed up and passed by the verifier. This is
the btf_struct_meta for the type being allocated. It would be needed
mostly for the offset array which is required for zero initializing
special fields while leaving the rest of storage in unitialized state.
It would also be needed in the next patch to perform proper destruction
of the object's special fields.
Under the hood, bpf_obj_new will call bpf_mem_alloc and bpf_mem_free,
using the any context BPF memory allocator introduced recently. To this
end, a global instance of the BPF memory allocator is initialized on
boot to be used for this purpose. This 'bpf_global_ma' serves all
allocations for bpf_obj_new. In the future, bpf_obj_new variants will
allow specifying a custom allocator.
Note that now that bpf_obj_new can be used to allocate objects that can
be linked to BPF linked list (when future linked list helpers are
available), we need to also free the elements using bpf_mem_free.
However, since the draining of elements is done outside the
bpf_spin_lock, we need to do migrate_disable around the call since
bpf_list_head_free can be called from map free path where migration is
enabled. Otherwise, when called from BPF programs migration is already
disabled.
A convenience macro is included in the bpf_experimental.h header to hide
over the ugly details of the implementation, leading to user code
looking similar to a language level extension which allocates and
constructs fields of a user type.
struct bar {
struct bpf_list_node node;
};
struct foo {
struct bpf_spin_lock lock;
struct bpf_list_head head __contains(bar, node);
};
void prog(void) {
struct foo *f;
f = bpf_obj_new(typeof(*f));
if (!f)
return;
...
}
A key piece of this story is still missing, i.e. the free function,
which will come in the next patch.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20221118015614.2013203-14-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2022-11-18 01:56:03 +00:00
|
|
|
#endif
|