ArmVirtPkg: introduce FdtSerialPortAddressLib

Introduce a new library class + instance for:

- collecting serial port base addresses from the device tree,

- collecting the /chosen stdout-path serial port base address from the
  device tree.

The logic is loosely based on the following functions:

- SerialPortGetBaseAddress()
  [ArmVirtPkg/Library/FdtPL011SerialPortLib/EarlyFdtPL011SerialPortLib.c]

- PlatformPeim() [ArmVirtPkg/Library/PlatformPeiLib/PlatformPeiLib.c]

- GetSerialConsolePortAddress()
  [ArmVirtPkg/Library/Fdt16550SerialPortHookLib/EarlyFdt16550SerialPortHookLib.c]

which are going to be converted to clients of the new library later.
Copyright notices from those other files are preserved.

The new library fixes the following warts, found by reading the existent
code:

- Neither of the three functions check whether the "reg" property exists.
  (This may be implicitly checked when they compare the property size to
  16.)

- GetSerialConsolePortAddress() uses ScanMem8() for locating a colon (":")
  node path separator in "stdout-path", when AsciiStrStr() could work just
  as fine. While ScanMem8() is likely faster, "stdout-path" is presumably
  very short, and ScanMem8() introduces an extra lib class dependency
  (namely BaseMemoryLib).

- If ScanMem8() fails to locate a colon in "stdout-path", then
  GetSerialConsolePortAddress() re-measures the length of the whole
  "stdout-path" property. This is conceptually (if not performance-wise)
  disturbing, because we know the whole size of the "stdout-path" property
  from the property lookup just before, so we only need to subtract the
  NUL-terminator for learning the length.

- GetSerialConsolePortAddress() does not check if the first (or only) node
  path inside the "stdout-path" property is empty. (Not a big deal, the
  subsequent alias resolution should simply fail.)

- GetSerialConsolePortAddress() does not verify if the node path retrieved
  (and potentially alias-resolved) from "stdout-path" can be located in
  the device tree; it assumes it.

- Code is duplicated (of course) between SerialPortGetBaseAddress() and
  PlatformPeim(), but more surprisingly, all three functions embed the
  same code for verifying the "status" property of the serial port node,
  and for checking and reading its "reg" property.

Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Leif Lindholm <quic_llindhol@quicinc.com>
Cc: Sami Mujawar <sami.mujawar@arm.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-Id: <20231008153912.175941-2-lersek@redhat.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=4577
[lersek@redhat.com: add TianoCore BZ reference]

0 files changed, 0 insertions, 0 deletions