In the current ongoing effort for sourcemod to fully support 64 bits, we are introducing "virtual address".
# Explanation
Because SourcePawn does not yet support a 64 bits-wide type it's been impossible for any plugins to hold addresses in regular 32-bits wide variable.
A first attempt at solving this issue was made in commit ce1a4dcac0 therein dubbed "PseudoAddress", however this turned out to be an unsatisfactory solution, as any 'high' address if offsetted could turn invalid (or outright be impossible to map).
This leaves us with three alternatives :
- New type
- Convert Address into a handle
- Virtual Address
A new type is the most destructive solution, as it entails breaking every single Address related method. While that solution is still not off the table, we're reserving it as the last attempt should this commit fail.
Converting into a handle type is a good compromise between a brand new type whilst also preserving the Address methods. However, this comes with two issues: the first being that you can no longer offset Address, the second is that we would require authors to free the handle type which will be very confusing. This will likely not be implemented.
# Virtual address
Under a reasonable assumption, we've noted that the average plugin is unlikely to play with more than 4 GB of memory; this shouldn't be too surprising as all valve games were once 32bits and therefore limited to 4GB. Assuming this stays mostly true and a plugin isn't interested with the mapped memory of lesser known modules (like soundlib or matlib), it is fair to assume plugins are unlikely to access more than 4GB of mapped memory. Working with this in mind, we map the memory the plugins are likely to access to our custom virtual address ranges (from 0 to 4Gb, the values of which can fit on 32bits variable). If any memory was missed and plugins were to try an access it later those ranges will be late-mapped to our virtual address ranges until we run out of them.
In order to use virtual addressing, whether on 32 bits or 64 bits. Plugins must now "#include <virtual_address>", as well as use the new SDKCall_VirtualAddress, SDKType_VirtualAddress, LoadAddressFromAddress & StoreAddressToAddress where it's appropriate to.
This brings in a few breaking changes.
One, INVALID_FUNCTION is now 0 instead of -1. This is long overdue.
Plugins should transparently work except in two cases:
1. Third-party extensions that have a hardcoded test for -1 will no
longer work. A new API has been provided for this,
GetFunctionByIdOrNull.
2. If a plugin "framework" uses INVALID_FUNCTION anywhere in its
exported API, then all plugins using that framework need to be
recompiled together, so they agree on the value of
INVALID_FUNCTION.
Hopefully the damage here is minimal. The core plugin version has been
bumped to 7 to try and limit conflicts.
Second, braceless functions are no longer supported. There wasn't really
any way around this and it's better to bite the bullet now. This affects
source compatibility, but not binary compatibility.
Third, the "using" keyword is no longer implemented. SourceMod now has a
Handle methodmap again. Plugins compiled against this new methodmap will
require a "Handle.~Handle" native, which 1.12 now provides.
* Harden plugin loading path requirements
Restrict loading of plugins to the `sourcemod/plugins` folder and require the `.smx` file extension.
Symlinks inside the `plugins` folder are fine.
This behavior was abused as part of justCTF 2020 in the PainterHell challenge by cypis. Thank you!
* Restrict extension loading to extensions folder
* Add NULL file extension check in LoadExtension
hi @KyleS
* Make mac/win lookups lowercase'd
* Revert #709 & 81042cc
* Adjust HashPolicy implementation across sourcemod
Basically, in order to implement our own (actual) hash policy in
`PluginSys.h`, we needed to remove the blanket implementation of `hash`
that was used before. Now, each policy must implement `hash` along with
`matches` in order to be used with `NameHashSet`. While this does force
us to change every implementation of policies across the entirety of
sourcemod, it allows core to use flexible implementations of `hash`.
* Remove logic duplication
* Improve lowercase checks
When unloading a plugin no plugin is allowed to execute code. There's
been some flawed logic in the scheduling, leading it to bail out if the
plugin was marked for scheduled unloading..
Also change the "error" message of `sm plugins unload` to something
nice.
1. Fixed OnPluginUnloaded not pairing if the plugin failed.
2. Unify error message handling in the second pass.
3. Do not add libraries if a plugin failed during OnPluginStart.
Note: this also ensures that library action callbacks are balanced (i.e., we do not notify
libraries are being dropped if they were never notified as being added).
