105
edits
Changes
Jump to navigation
Jump to search
mmap() bug: fixed, but kerndat need some care
Line 69:
Line 69:
Presumply, because of using syscall() and memcpy() from Glibc, instead of raw pie-helpers.+
== Needs to be done (TODO) ==
== Needs to be done (TODO) ==
−=== Bug with mmaping over 4Gb ===
+=== Kerndat test: check presence of the bug with mmaping over 4Gb ===
As 32-bit application is restored from 64-bit CRIU, some task's properties that were filled on <code>exec()</code> are left, which is quite unusual for 32-bit task. One of the things, left from 64-bit binary is precalculated <code>mmap_base</code> which is used to find task's top/bottom address limit during <code>mmap()</code> syscall. That means that compat <code>sys_mmap()</code> may map page over 4Gb address and return 4-byte pointer to low bytes of address. Looks like no one has used compatible mmap in 64-bit binary. Results in broken mmap in restored 32-bit application, which can map vma over 4Gb.
As 32-bit application is restored from 64-bit CRIU, some task's properties that were filled on <code>exec()</code> are left, which is quite unusual for 32-bit task. One of the things, left from 64-bit binary is precalculated <code>mmap_base</code> which is used to find task's top/bottom address limit during <code>mmap()</code> syscall. That means that compat <code>sys_mmap()</code> may map page over 4Gb address and return 4-byte pointer to low bytes of address. Looks like no one has used compatible mmap in 64-bit binary. Results in broken mmap in restored 32-bit application, which can map vma over 4Gb.
−See [[Upstream kernel commits]]. The kerndat test for 32-bit C/R will be reworked to check if the bug present in kernel (which is not nice thing, but ok).
+See [[Upstream kernel commits]], the bug is fixed in linux-next, but the kerndat criu test should be reworked to check presence of this bug, not only arch_prctl() for 32-bit vdso.
=== Issue with compat vdso helper ===
=== Issue with compat vdso helper ===
Rarely it segfaults: [https://github.com/xemul/criu/issues/290 issue], [https://github.com/xemul/criu/issues/273 reproduce].
Rarely it segfaults: [https://github.com/xemul/criu/issues/290 issue], [https://github.com/xemul/criu/issues/273 reproduce].
−Presumably, because of using syscall() and memcpy() from Glibc, instead of raw pie-helpers at the moment of unmapped vdso/vvar vmas.
Hard to reproduce.
Hard to reproduce.
Line 94:
Line 94:
| vdso01 || need to add 32-bit vDSO symbols: __kernel_* (and check them with handlers in test)
| vdso01 || need to add 32-bit vDSO symbols: __kernel_* (and check them with handlers in test)
|-
|-
−| vdso01 || fails at fault-inject trampoline testing: correct jump trampolines [https://lists.openvz.org/pipermail/criu/2017-March/036187.html Fixed]
+| vdso01 || fails at fault-inject trampoline testing: correct jump trampolines [https://lists.openvz.org/pipermail/criu/2017-March/036187.html Fixed in this set]
|}
|}