Kernel/Filesystem+Tests: Fix RAMFS block count reporting, introduce a test

[sdomi]

I tried making the test for this as generic as possible, so we can extend it onto more FSes (once we start testing those more widely in the future). Currently it runs against RAMFS itself and Ext2FS, because that was the easiest to set up.

L21 from TestFileSystemReportedSize.cpp isn't an EXPECT_EQ due to the macro complaining about not being able to compare between two types; maybe this is something we should look into fixing in the future?

[timschumi]

L21 from TestFileSystemReportedSize.cpp isn't an EXPECT_EQ due to the macro complaining about not being able to compare between two types; maybe this is something we should look into fixing in the future?

I'm not sure if this is fixable unless we want to go into templated functions instead of macros. We assign the left-hand side and the right-hand side to auto variables first, and C++ isn't smart enough to restrict the types based on the following comparison (instead taking the type from the immediate).

In this particular case, EXPECT_EQ(st.st_blocks, 0ull); should make it work.

EDIT: Remainder of the review is pending, once I'm back at my PC.

[supercomputer7]

Does it fix the issue actually? I don't see where we set the block size for RAMFS so it will essentially multiply the block count by 0, isn't it correct?

[timschumi]

Block size is assumed to be 512/1024/"whatever the user puts in" by du. I don't think POSIX has a facility to retrieve the filesystem block size?

[timschumi]

Addendum: Ah, yeah, there is st_blksize. If we have that implemented then we should make sure that it's set.

[sdomi]

st_blksize is being set by two classes down the line, in BlockBasedFileSystem.h (precisely here). I checked, and for our purposes it's always 512. Is this enough, or should we for some reason set it again in this Inode.cpp?

[supercomputer7]

RAMFS is not inheriting from BlockBasedFileSystem, so setting this there is probably not affecting us anyway.

Each data block is using a 128 KiB space, so 512 bytes is probably incorrect anyway.

serenity/Kernel/FileSystem/RAMFS/Inode.h

Lines 69 to 75 in ad4bd36

	struct DataBlock {
	public:
	using List = Vector<OwnPtr<DataBlock>>;
	static ErrorOr<NonnullOwnPtr<DataBlock>> create();
	constexpr static size_t block_size = 128 * KiB;

In #19867, I actually reduced this to 4 KiB. But 128 KiB is way too big to do anything useful with it.
Maybe change the block size to PAGE_SIZE:

constexpr static size_t data_block_size = PAGE_SIZE;

and then change st_blksize by setting it to that size as well.

[sdomi]

oops, mb. I'm not sure where it gets the 512 from then (I did verify that it does)

will push a fix later