On 24-Apr-20 11:50 AM, Li Feng wrote:
Currently, even though memory is mapped with PROT_NONE, this does not
cause it to be excluded from core dumps. This is counter-productive,
because in a lot of cases, this memory will go unused (e.g. when the
memory subsystem preallocates VA space but hasn't yet mapped physical
pages into it).
Use `madvise()` call with MADV_DONTDUMP parameter to exclude the
unmapped memory from being dumped.
Signed-off-by: Li Feng <fen...@smartx.com>
---
V2:
- add support for freebsd.
- when free_seg is called, mark the memory MADV_DONTDUMP.
- when alloc_seg is called, mark the memory MADV_DODUMP.
Isn't this v3 now?
lib/librte_eal/common/eal_common_memory.c | 26 ++++++++++++++++++++++++++
lib/librte_eal/linux/eal_memalloc.c | 6 ++++++
2 files changed, 32 insertions(+)
diff --git a/lib/librte_eal/common/eal_common_memory.c
b/lib/librte_eal/common/eal_common_memory.c
index cc7d54e0c..83be94a20 100644
--- a/lib/librte_eal/common/eal_common_memory.c
+++ b/lib/librte_eal/common/eal_common_memory.c
@@ -177,6 +177,32 @@ eal_get_virtual_area(void *requested_addr, size_t *size,
after_len = RTE_PTR_DIFF(map_end, aligned_end);
if (after_len > 0)
munmap(aligned_end, after_len);
+
+ /*
+ * Exclude this pages from a core dump.
+ */
+#ifdef RTE_EXEC_ENV_LINUX
+ if (madvise(aligned_addr, *size, MADV_DONTDUMP) != 0)
+ RTE_LOG(DEBUG, EAL, "madvise failed: %s\n",
+ strerror(errno));
+#elif RTE_EXEC_ENV_FREEBSD
+ if (madvise(aligned_addr, *size, MADV_NOCORE) != 0)
+ RTE_LOG(DEBUG, EAL, "madvise failed: %s\n",
+ strerror(errno));
+#endif
+ } else {
+ /*
+ * Exclude this pages from a core dump.
+ */
+#ifdef RTE_EXEC_ENV_LINUX
+ if (madvise(mapped_addr, map_sz, MADV_DONTDUMP) != 0)
+ RTE_LOG(DEBUG, EAL, "madvise failed: %s\n",
+ strerror(errno));
+#elif RTE_EXEC_ENV_FREEBSD
+ if (madvise(mapped_addr, map_sz, MADV_NOCORE) != 0)
+ RTE_LOG(DEBUG, EAL, "madvise failed: %s\n",
+ strerror(errno));
+#endif
}
DRY (Don't Repeat Yourself) :) Probably easier to do "if (!unmap) ..."
than putting this into two places.
return aligned_addr;
diff --git a/lib/librte_eal/linux/eal_memalloc.c
b/lib/librte_eal/linux/eal_memalloc.c
index af6d0d023..9d2a6fc6f 100644
--- a/lib/librte_eal/linux/eal_memalloc.c
+++ b/lib/librte_eal/linux/eal_memalloc.c
@@ -571,6 +571,9 @@ alloc_seg(struct rte_memseg *ms, void *addr, int socket_id,
goto resized;
}
+ if (madvise(addr, alloc_sz, MADV_DODUMP) != 0)
+ RTE_LOG(DEBUG, EAL, "madvise failed: %s\n", strerror(errno));
+
See discussion for v2, i believe this is unnecessary.
/* In linux, hugetlb limitations, like cgroup, are
* enforced at fault time instead of mmap(), even
* with the option of MAP_POPULATE. Kernel will send
@@ -687,6 +690,9 @@ free_seg(struct rte_memseg *ms, struct hugepage_info *hi,
return -1;
}
+ if (madvise(ms->addr, ms->len, MADV_DONTDUMP) != 0)
+ RTE_LOG(DEBUG, EAL, "madvise failed: %s\n", strerror(errno));
+
exit_early = false;
/* if we're using anonymous hugepages, nothing to be done */
--
Thanks,
Anatoly
