https://bugs.kde.org/show_bug.cgi?id=526974
--- Comment #3 from Martin Cermak <[email protected]> --- Created attachment 197179 --> https://bugs.kde.org/attachment.cgi?id=197179&action=edit proposed patch The LTP update was accepted and merged, great. But there still is a problem that mremap03 behaves differently with vs. without Valgrind. I'm attaching a proposed patch that fixes this on the Valgrind side. The LTP mremap03 essentially creates 1page VMA via tst_get_bad_addr(). Later on it calls mremap(bad_addr, memsize, ...) where memsize is 1000 times larger than that. That's an invalid request, which in the kernel seems to be handled in linux/mm/mremap.c:1850: 1836 /* 1837 * We can't remap across the end of VMAs, as another VMA may be 1838 * adjacent: 1839 * 1840 * addr vma->vm_end 1841 * |-----.----------| 1842 * | . | 1843 * |-----.----------| 1844 * .<--------->xxx> 1845 * old_len 1846 * 1847 * We also require that vma->vm_start <= addr < vma->vm_end. 1848 */ 1849 if (old_len > vma->vm_end - addr) 1850 return -EFAULT; This also matches what the testcase expects (EFAULT). Testing the attached patch now... -- You are receiving this mail because: You are watching all bug changes.
