Thanks to Pieter Wuille for most of the work on this commit.
I did not fixup the overhaul commit, because a rebase conflicted
with "remove fields of ser_streamplaceholder".
I prefer not to risk making a mistake while resolving it.
The implementation of each class' serialization/deserialization is no longer
passed within a macro. The implementation now lies within a template of form:
template <typename T, typename Stream, typename Operation>
inline static size_t SerializationOp(T thisPtr, Stream& s, Operation ser_action, int nType, int nVersion) {
size_t nSerSize = 0;
/* CODE */
return nSerSize;
}
In cases when codepath should depend on whether or not we are just deserializing
(old fGetSize, fWrite, fRead flags) an additional clause can be used:
bool fRead = boost::is_same<Operation, CSerActionUnserialize>();
The IMPLEMENT_SERIALIZE macro will now be a freestanding clause added within
class' body (similiar to Qt's Q_OBJECT) to implement GetSerializeSize,
Serialize and Unserialize. These are now wrappers around
the "SerializationOp" template.
- ensures a consistent usage in header files
- also add a blank line after the copyright header where missing
- also remove orphan new-lines at the end of some files
Use misc methods of avoiding unnecesary header includes.
Replace int typedefs with int##_t from stdint.h.
Replace PRI64[xdu] with PRI[xdu]64 from inttypes.h.
Normalize QT_VERSION ifs where possible.
Resolve some indirect dependencies as direct ones.
Remove extern declarations from .cpp files.
This ensures the allocator is ready no matter when it's needed (as
some STL implementations allocate in constructors -- i.e., MSVC's STL
in debug builds).
Using boost::call_once to guarantee thread-safe static initialization.
Adding some comments describing why the change was made.
Addressing deinitialization of the LockedPageManager object
by initializing it in a local static initializer and adding
an assert in the base's destructor.
As memset() can be optimized out by a compiler it should not be used in
privacy/security relevant code parts. OpenSSL provides the safe
OPENSSL_cleanse() function in crypto.h, which perfectly does the job of
clean and overwrite data.
For details see: http://www.viva64.com/en/b/0178/
- change memset() to OPENSSL_cleanse() where appropriate
- change a hard-coded number from netbase.cpp into a sizeof()
Replace direct calls to mlock.
Also, change the class to lock the memory areas in the constructor and unlock them again in the destructor. This makes sure that locked pages won't leak.
This commit removes the dependency of serialize.h on PROTOCOL_VERSION,
and makes this parameter required instead of implicit. This is much saner,
as it makes the places where changing a version number can have an
influence obvious.
SecureString is identical to std::string except with secure_allocator
substituting for std::allocator. This makes casting between them
impossible, so converting between the two at API boundaries requires
calling ::c_str() for now.
This commit adds support for ckeys, or enCrypted private keys, to the wallet.
All keys are stored in memory in their encrypted form and thus the passphrase
is required from the user to spend coins, or to create new addresses.
Keys are encrypted with AES-256-CBC using OpenSSL's EVP library. The key is
calculated via EVP_BytesToKey using SHA512 with (by default) 25000 rounds and
a random salt.
By default, the user's wallet remains unencrypted until they call the RPC
command encryptwallet <passphrase> or, from the GUI menu, Options->
Encrypt Wallet.
When the user is attempting to call RPC functions which require the password
to unlock the wallet, an error will be returned unless they call
walletpassphrase <passphrase> <time to keep key in memory> first.
A keypoolrefill command has been added which tops up the users keypool
(requiring the passphrase via walletpassphrase first).
keypoolsize has been added to the output of getinfo to show the user the
number of keys left before they need to specify their passphrase (and call
keypoolrefill).
Note that walletpassphrase will automatically fill keypool in a separate
thread which it spawns when the passphrase is set. This could cause some
delays in other threads waiting for locks on the wallet passphrase, including
one which could cause the passphrase to be stored longer than expected,
however it will not allow the passphrase to be used longer than expected as
ThreadCleanWalletPassphrase will attempt to get a lock on the key as soon
as the specified lock time has arrived.
When the keypool runs out (and wallet is locked) GetOrReuseKeyFromPool
returns vchDefaultKey, meaning miners may start to generate many blocks to
vchDefaultKey instead of a new key each time.
A walletpassphrasechange <oldpassphrase> <newpassphrase> has been added to
allow the user to change their password via RPC.
Whenever keying material (unencrypted private keys, the user's passphrase,
the wallet's AES key) is stored unencrypted in memory, any reasonable attempt
is made to mlock/VirtualLock that memory before storing the keying material.
This is not true in several (commented) cases where mlock/VirtualLocking the
memory is not possible.
Although encryption of private keys in memory can be very useful on desktop
systems (as some small amount of protection against stupid viruses), on an
RPC server, the password is entered fairly insecurely. Thus, the only main
advantage encryption has for RPC servers is for RPC servers that do not spend
coins, except in rare cases, eg. a webserver of a merchant which only receives
payment except for cases of manual intervention.
Thanks to jgarzik for the original patch and sipa, gmaxwell and many others
for all their input.
Conflicts:
src/wallet.cpp
Pieter Wuille