bytes_ostream is an incremental builder for a discontiguous byte container.
managed_bytes is a non-incremental (size must be known up front) byte
container, that is also compatible with LSA. So far, conversion between
them involves copying. This is unfortunate, since query_result is generated
as a bytes_ostream, but is later converted to managed_bytes (today, this
is done in cql3::expr::get_non_pk_values() and
compound_view_wrapper::explode(). If the two types could be made compatible,
we could use managed_bytes_view instead of creating new objects and avoid
a copy. It's also nicer to have one less vocabulary type.
This patch makes bytes_ostream use managed_bytes' internal representation
(blob_storage instead of bytes_ostream::chunk) and provides a conversion
to managed_bytes. All bytes_ostream users are left in place, but the goal
is to make bytes_ostream a write-only type with the only observer a conversion
to managed_bytes.
It turns out to be relatively simple. The internal representations were
already similar. I made blob_storage::ref_type self-initializing to
reduce churn (good practice anyway) and added a private constructor
to managed_bytes for the conversion.
Note that bytes_ostream can only be used to construct a non-LSA managed_bytes,
but LSA uses of managed_bytes are very strictly controlled (the entry
points to memtable and cache) so that's not a problem.
A unit test is added.
Closes#10986
Instead of lengthy blurbs, switch to single-line, machine-readable
standardized (https://spdx.dev) license identifiers. The Linux kernel
switched long ago, so there is strong precedent.
Three cases are handled: AGPL-only, Apache-only, and dual licensed.
For the latter case, I chose (AGPL-3.0-or-later and Apache-2.0),
reasoning that our changes are extensive enough to apply our license.
The changes we applied mechanically with a script, except to
licenses/README.md.
Closes#9937
managed_bytes_basic_view is a template with a constructor that
converts from one instantiation of the template to another.
Unfortunately when gcc encounters the associated constraint, it
instantiates the template which forces it to evaluate the constraint
again, sending it into a loop.
Fix that by making the converting constructor a template itself,
delaying instantiation. The constraint is strengthened so the set
of types on which the constructor works is unchanged.
The range_tombstone_list's method is at the top of the
stack of calls each not throwing anything, so do the
deep-dive noexcept marking.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Yet another patch preventing potentially large allocations.
Currently, collection_mutation{_view,}_description linearize each collection
value during deserialization. It's not unthinkable that a user adds a
large element to a list or a map, so let's avoid that.
This patch removes the dependency on linearizing_input_stream, which does not
provide a way to read fragmented subbuffers, and replaces it with a new
helper, which does. (Extending linearizing_input_stream is not viable without
rewriting it completely).
Only linearization of collection values is corrected in this patch.
Collection keys are still linearized. Storing them in managed_bytes is likely
to be more harmful than helpful, because large map keys are extremely unlikely,
and UUIDs, which are used as keys in lists, do not fit into manages_bytes's
small value optimization, so this would incure an extra allocation for every
list element.
Note: this patch leaves utils/linearizing_input_stream.hh unused.
Refs: #8120Closes#8690
Usually lsa allocation is performed with the construct() helper that
allocates a sizeof(T) slot and constructs it in place. Some rare
objects have dynamic size, so they are created by alloc()ating a
slot of some specific size and (!) must provide the correct overload
of size_for_allocation_strategy that reports back the relevant
storage size.
This "must provide" is not enforced, if missed a default sizer would
be instantiated, but won't work properly. This patch makes all users
of alloc() conform to DynamicObject concept which requires the
presense of .storage_size() method to tell that size.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Todays alloc() accepts migrate-fn, size and alignment. All the callers
don't really need to provide anything special for the migrate-fn and
are just happy with default alignof() for alignment. The simplification
is in providing alloc() that only accepts size arg and does the rest
itself.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Just for convenience. We will use it in an upcoming patch where we switch
the inner representation of cql3::raw_value from bytes to managed_bytes, and we
will want to construct managed_bytes from fragmented_temporary_buffer::view.
We will need them to replace bytes with managed_bytes in some places in an
upcoming patch.
The change to configure.py is necessary because opearator<< links to to_hex
in bytes.cc.
managed_bytes has a small overhead per each fragment. Due to that, managed_bytes
containing the same data can have different total memory usage in different
allocators. The smaller the preferred max allocation size setting is, the more
fragments are needed and the greater total per-fragment overhead is.
In particular, managed_bytes allocated in the LSA could grow in
memory usage when copied to the standard allocator, if the standard allocator
had a preferred max allocation setting smaller than the LSA.
partition_snapshot_accounter calculates the amount of memory used by
mutation fragments in the memtable (where they are allocated with LSA) based
on the memory usage after they are copied to the standard allocator.
This could result in an overestimation, as explained above.
But partition_snapshot_accounter must not overestimate the amount of freed
memory, as doing otherwise might result in OOM situations.
This patch prevents the overaccounting by adding minimal_external_memory_usage():
a new version of external_memory_usage(), which ignores allocator-dependent
overhead. In particular, it includes the per-fragment overhead in managed_bytes
only once, no matter how many fragments there are.
Remove the following bits of `managed_bytes` since they are unused:
* `with_linearized_managed_bytes` function template
* `linearization_context_guard` RAII wrapper class for managing
`linearization_context` instances.
* `do_linearize` function
* `linearization_context` class
Since there is no more public or private methods in `managed_class`
to linearize the value except for explicit `with_linearized()`,
which doesn't use any of aforementioned parts, we can safely remove
these.
Signed-off-by: Pavel Solodovnikov <pa.solodovnikov@scylladb.com>
This operator has a single purpose: an easier port of legacy_compound_view
from bytes_view to managed_bytes_view.
It is inefficient and should be removed as soon as legacy_compound_view stops
using operator[].
managed_bytes_view is a non-owning view into managed_bytes.
It can also be implicitly constructed from bytes_view.
It conforms to the FragmentedView concept and is mainly used through that
interface.
It will be used as a replacement for bytes_view occurrences currently
obtained by linearizing managed_bytes.
Conversions from views to owners have no business being implicit.
Besides, they would also cause various ambiguity problems when adding
managed_bytes_view.
This is a temporary scaffold for weaning ourselves off
linearization. It differs from with_linearized_managed_bytes in
that it does not rely on the environment (linearization_context)
and so is easier to remove.
We use managed_bytes::data() in a few places when we know the
data is non-fragmented (such as when the small buffer optimization
is in use). We'd like to remove managed_bytes::data() as linearization
is bad, so in preparation for that, replace internal uses of data()
with the equivalent direct access.
do_linearize() is an impure function as it changes state
in linearization_context. Extract the pure parts into a new
do_linearize_pure(). This will be used to linearize managed_bytes
without a linearization_context, during the transition period where
fragmented and non-fragmented values coexist.
If external is true, _u.ptr is not null. An empty managed_bytes uses
the internal representation.
The current code looks scary, since it seems possible that backref
would still point to the old location, which would invite corruption
when the reclaimer runs.
Signed-off-by: Rafael Ávila de Espíndola <espindola@scylladb.com>
Reviewed-by: Benny Halevy <bhalevy@scylladb.com>
Message-Id: <20200716233124.521796-1-espindola@scylladb.com>
The compilation fails on -Warray-bounds, even though the branch is never taken:
inlined from ‘managed_bytes::managed_bytes(bytes_view)’ at ./utils/managed_bytes.hh:195:22,
inlined from ‘managed_bytes::managed_bytes(const bytes&)’ at ./utils/managed_bytes.hh:162:77,
inlined from ‘dht::token dht::bytes_to_token(bytes)’ at dht/random_partitioner.cc:68:57,
inlined from ‘dht::token dht::random_partitioner::get_token(bytes)’ at dht/random_partitioner.cc:85:39:
/usr/include/c++/8/bits/stl_algobase.h:368:23: error: ‘void* __builtin_memmove(void*, const void*, long unsigned int)’ offset 16 from the object at ‘<anonymous>’ is out of the bounds of referenced subobject ‘managed_bytes::small_blob::data’ with type ‘signed char [15]’ at offset 0 [-Werror=array-bounds]
__builtin_memmove(__result, __first, sizeof(_Tp) * _Num);
~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Work around by disabling the diagnostic locally.
Message-Id: <1547205350-30225-1-git-send-email-tgrabiec@scylladb.com>
Since linearization_context is thread_local every time it is accessed
the compiler needs to emit code that checks if it was already
constructed and does so if it wasn't. Moreover, upon leaving the context
from the outermost scope the map needs to be cleared.
All these operations impose some performance overhead and aren't really
necessary if no buffers were linearised (the expected case). This patch
rearranges the code so that lineatization_context is trivially
constructible and the map is cleared only if it was modified.
Because of the small size optimization, not all copies will call the
allocator, so allocation failure injection may miss this site if the
value is not large enough. Make the testing more effective by marking
this place explicitly as an allocation point.
While blob_storage is marked as an unaligned type, the back references also
point to an unaligned type (a pointer to blob_storage), since a back
reference can live in a blob_storage. This triggers errors from zapcc/clang 4.
Fix by creating a type for the reference, which is marked as unaligned.
Message-Id: <20170502071404.507-1-avi@scylladb.com>
Every lsa-allocated object is prefixed by a header that contains information
needed to free or migrate it. This includes its size (for freeing) and
an 8-byte migrator (for migrating). Together with some flags, the overhead
is 14 bytes (16 bytes if the default alignment is used).
This patch reduces the header size to 1 byte (8 bytes if the default alignment
is used). It uses the following techniques:
- ULEB128-like encoding (actually more like ULEB64) so a live object's header
can typically be stored using 1 byte
- indirection, so that migrators can be encoded in a small index pointing
to a migrator table, rather than using an 8-byte pointer; this exploits
the fact that only a small number of types are stored in LSA
- moving the responsibility for determining an object's size to its
migrator, rather than storing it in the header; this exploits the fact
that the migrator stores type information, and object size is in fact
information about the type
The patch improves the results of memory_footprint_test as following:
Before:
- in cache: 976
- in memtable: 947
After:
mutation footprint:
- in cache: 880
- in memtable: 858
A reduction of about 10%. Further reductions are possible by reducing the
alignment of lsa objects.
logalloc_test was adjusted to free more objects, since with the lower
footprint, rounding errors (to full segments) are different and caused
false errors to be detected.
Missing: adjustments to scylla-gdb.py; will be done after we agree on the
new descriptor's format.
Renaming the function to external_memory_usage() makes it clear that
sizeof(T) is not included, something that was a source of confusion in
the past.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
It is needed for noexcept destruction, which we need for exception
safety in higher layers.
According to [1], erase() only throws if key comparison throws, and in
our case it doesn't.
[1] http://en.cppreference.com/w/cpp/container/unordered_map/erase
Fixes#865
(Some) gcc 5 (5.3.0 for me) on ubuntu will generate errors on
compilation of this code (compiling logalloc_test). The memcpy
to inline storage seems to confuse the compiler.
Simply change to std::copy, which shuts the compiler up.
Any decent stl should convert primitive std::copy to memcpy
anyway, but since it is also the inline (small storage),
it should not matter which way.
Message-Id: <1456931988-5876-4-git-send-email-calle@scylladb.com>
A large managed_bytes blob can be scattered in lsa memory. Usually this is
fine, but someone we want to examine it in place without copying it out, but
using contiguous iterators for efficiency.
For this use case, introduce with_linearized_managed_bytes(Func),
which runs a function in a "linearization context". Within the linearization
context, reads of managed_bytes object will see temporarily linearized copies
instead of scattered data.
blob_storage defined with attribute packed which makes its alignment
requirement equal 1. This means that its members may be unaligned.
GCC is obviously aware of that and will generate appropriate code
(and not generate ubsan checks). However, there are few places where
members of blob_storage are accessed via pointers, these have to be
wrapped by unaligned_cast<> to let the compiler know that the location
pointed to may be not aligned properly.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Instead of allocating a single blob_storage, chain multiple blob_storage
objects in a list, each limited not to exceed the allocation_strategy's
max_preferred_allocation_size. This allows lsa to allocate each blob_storage
object as an lsa managed object that can be migrated in memory.
Also provide linearize()/scatter() methods that can be used to temporarily
consolidate the storage into a single blob_storage. This makes the data
contiguous, so we can use a regular bytes_view to examine it.
