What does an equals sign = on the right side of a LDR instruction in ARM mean?

Question

Been googling this for a while but i can't find any documentation relating to this. I've been trying to learn ARM and have been looking at the compiled ARM assembly code for a simple calculator.c program i wrote in order to see if I could understand what was going on. The thing I keep seeing is instructions like these:

LDR     R3, =__stack_chk_guard__GLIBC_2.4

or

LDR     R0, =aEnterOperator ; "Enter operator: "

or

LDR     R0, =aSIsNotAValidOp ; "%s is not a valid operator.  Enter +, -"

Note: the stuff after the semicolons is just the auto-comments added by IDA.

My question is, what does the '=' on the right side of these LDRs mean? In the first case, it seems to be some tag indicating the loading of a library; in the second and third cases, '=a' seems to be prefacing a printf. I'm just not quite sure to make of this, since I can't find anything about this syntax for LDR in the documentation. Can someone help me understand this? Thank you!

Shawannashawl

Which assembler do you use? Did you read the documentation of the Assembler? – Purposeful

Related: #17215462 – Bonnett

Answer 1

The use of an equals sign (=) at the start of the second operand of the LDR instruction indicates the use of the LDR pseudo-instruction. This pseuo-instruction is used to load an arbitrary 32-bit constant value into a register with a single instruction despite the fact that the ARM instruction set only supports immediate values in a much smaller range.

If the value after the = is known by the assembler and fits in with the allowed range of an immediate value for the MOV or MVN instruction then a MOV or MVN instruction is generated. Otherwise the constant value is put into the literal pool, and a PC-relative LDR instruction is used to load the value into the register.

If Ida is generating these LDR= instructions when dissassembling code then it must have detected that the assembler or compiler chose the second option when generating the code you're looking at. The actual instruction is something like LDR R0, loc_1234567 (or more accurately something like LDR R0, [PC, #-1234]) and Ida is looking up the value in the literal pool at loc_1234567 for you.

Emission

Answer 2

= is usually suffixed by an immediate constant and instructs the assembler to put the constant into a nearby literal pool and generate a pc relative memory operand to load it. This is useful since the ARM instruction format doesn't have enough space to store a full 32 bit constant. Loading constants that cannot be encoded in 8 bits (I think) plus a shift from a nearby literal pool is an effective and efficient way to circumvent this problem.

Dwelt