x86/e820/32: Fix e820_search_gap() error handling on x86-32

GCC correctly points out that on 32-bit kernels, e820_search_gap()
not finding a start now leads to pci_mem_start ('gapstart') being set to an
uninitialized value:

  arch/x86/kernel/e820.c: In function 'e820_setup_gap':
  arch/x86/kernel/e820.c:641:16: error: 'gapstart' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This restores the behavior from before this cleanup:

  b4ed1d15 ("x86/e820: Make e820_search_gap() static and remove unused variables")

... defaulting to address 0x10000000 if nothing was found.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Toshi Kani <toshi.kani@hpe.com>
Cc: Wei Yang <richard.weiyang@gmail.com>
Fixes: b4ed1d15 ("x86/e820: Make e820_search_gap() static and remove unused variables")
Link: http://lkml.kernel.org/r/20170111144926.695369-1-arnd@arndb.deSigned-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/kernel/e820.c

@@ -626,14 +626,16 @@ __init void e820_setup_gap(void)
 	gapsize = 0x400000;
 	found  = e820_search_gap(&gapstart, &gapsize);
 
-#ifdef CONFIG_X86_64
 	if (!found) {
+#ifdef CONFIG_X86_64
 		gapstart = (max_pfn << PAGE_SHIFT) + 1024*1024;
 		printk(KERN_ERR
 	"e820: cannot find a gap in the 32bit address range\n"
 	"e820: PCI devices with unassigned 32bit BARs may break!\n");
-	}
+#else
+		gapstart = 0x10000000;
 #endif
+	}
 
 	/*
 	 * e820_reserve_resources_late protect stolen RAM already