Assembly executing a long jump with an offset with different syntax

A few things. The AT&T syntax for a far jump is:

jmp $0x08,$flush2 

The label in this case needs to be preceded by a $. Immediate values like 0x08 also need a $. This line doesn't do what you think it does:

mov %ax, 0x10

Important thing about AT&T syntax is that unlike Intel Syntax the operands are reversed. Source operand is first and destination operation is after. Secondly, immediate values in AT&T syntax on x86/x86-64 need to have a $ sign prepended to them or they are actually treated as a memory operand. Your instruction actually moved the 16-bit contents of AX to the memory address 0x00000010 which is not what you intended. What you wanted was:

mov $0x10, %ax

This moves the immediate value 0x10 to AX. The problem with operands being reversed also applies to all your lines like:

mov %ds, %ax

Should be:

mov %ax, %ds

I usually prefer calling your function load_gdt. I'm usually a fan of passing the segment values (CS and DS) and the address of the GDTR with code like:

load_gdt:
    mov 4(%esp), %edx    # EDX is 1st argument - GDT record pointer
    mov 8(%esp), %eax    # EAX is 2nd argument - Data Selector
    lgdt (%edx)          # Load GDT with GDT record pointer passed as 1st argument
    mov %eax, %ds        # Reload all the data descriptors with Data selector (2nd arg)
    mov %eax, %es
    mov %eax, %gs
    mov %eax, %fs
    mov %eax, %ss

    pushl 12(%esp)      # Create FAR pointer on stack using Code selector (3rd argument)
    push $.setcs         # Offset of FAR JMP will be setcs label below
    ljmp *(%esp)        # Do the FAR JMP to next instruction to set CS with Code selector,
                        #    and set the EIP (instruction pointer) to offset of setcs
.setcs:
    add $8, %esp        # Restore stack (remove 2 DWORD values we put on stack to
                        #     create FAR Pointer)
    ret

The C prototype would be something like:

void load_gdt(struct gdt_ptr *gdt_ptr, unsigned int data_sel, unsigned int code_sel);

If you want to use GNU assembler with a variant of Intel Syntax you could try adding this directive to the top of all your assembly files:

.intel_syntax noprefix

In addition to Michael's answer, which definitely out-informs mine, this would be my translation:

.global gdt_flush

gdt_flush:
   movl 4(%esp),%eax
   lgdt (%eax)

   movw $0x10, %ax
   movw %ax, %ds
   movw %ax, %es
   movw %ax, %fs
   movw %ax, %gs
   movw %ax, %ss
   jmp  $0x08,$flush

flush:
   ret