fs/proc/kcore: drop KCORE_REMAP and KCORE_OTHER
authorDavid Hildenbrand <david@redhat.com>
Thu, 1 Jul 2021 01:50:03 +0000 (18:50 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 1 Jul 2021 03:47:28 +0000 (20:47 -0700)
Patch series "fs/proc/kcore: don't read offline sections, logically offline pages and hwpoisoned pages", v3.

Looking for places where the kernel might unconditionally read
PageOffline() pages, I stumbled over /proc/kcore; turns out /proc/kcore
needs some more love to not touch some other pages we really don't want to
read -- i.e., hwpoisoned ones.

Examples for PageOffline() pages are pages inflated in a balloon, memory
unplugged via virtio-mem, and partially-present sections in memory added
by the Hyper-V balloon.

When reading pages inflated in a balloon, we essentially produce
unnecessary load in the hypervisor; holes in partially present sections in
case of Hyper-V are not accessible and already were a problem for
/proc/vmcore, fixed in makedumpfile by detecting PageOffline() pages.  In
the future, virtio-mem might disallow reading unplugged memory -- marked
as PageOffline() -- in some environments, resulting in undefined behavior
when accessed; therefore, I'm trying to identify and rework all these
(corner) cases.

With this series, there is really only access via /dev/mem, /proc/vmcore
and kdb left after I ripped out /dev/kmem.  kdb is an advanced corner-case
use case -- we won't care for now if someone explicitly tries to do nasty
things by reading from/writing to physical addresses we better not touch.
/dev/mem is a use case we won't support for virtio-mem, at least for now,
so we'll simply disallow mapping any virtio-mem memory via /dev/mem next.
/proc/vmcore is really only a problem when dumping the old kernel via
something that's not makedumpfile (read: basically never), however, we'll
try sanitizing that as well in the second kernel in the future.

Tested via kcore_dump:
https://github.com/schlafwandler/kcore_dump

This patch (of 6):

Commit db779ef67ffe ("proc/kcore: Remove unused kclist_add_remap()")
removed the last user of KCORE_REMAP.

Commit 595dd46ebfc1 ("vfs/proc/kcore, x86/mm/kcore: Fix SMAP fault when
dumping vsyscall user page") removed the last user of KCORE_OTHER.

Let's drop both types.  While at it, also drop vaddr in "struct
kcore_list", used by KCORE_REMAP only.

Link: https://lkml.kernel.org/r/20210526093041.8800-1-david@redhat.com
Link: https://lkml.kernel.org/r/20210526093041.8800-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Aili Yao <yaoaili@kingsoft.com>
Cc: Jiri Bohac <jbohac@suse.cz>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/kcore.c
include/linux/kcore.h

index 4d2e64e..09f77d3 100644 (file)
@@ -380,11 +380,8 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos)
                        phdr->p_type = PT_LOAD;
                        phdr->p_flags = PF_R | PF_W | PF_X;
                        phdr->p_offset = kc_vaddr_to_offset(m->addr) + data_offset;
-                       if (m->type == KCORE_REMAP)
-                               phdr->p_vaddr = (size_t)m->vaddr;
-                       else
-                               phdr->p_vaddr = (size_t)m->addr;
-                       if (m->type == KCORE_RAM || m->type == KCORE_REMAP)
+                       phdr->p_vaddr = (size_t)m->addr;
+                       if (m->type == KCORE_RAM)
                                phdr->p_paddr = __pa(m->addr);
                        else if (m->type == KCORE_TEXT)
                                phdr->p_paddr = __pa_symbol(m->addr);
index da676cd..86c0f1d 100644 (file)
@@ -11,14 +11,11 @@ enum kcore_type {
        KCORE_RAM,
        KCORE_VMEMMAP,
        KCORE_USER,
-       KCORE_OTHER,
-       KCORE_REMAP,
 };
 
 struct kcore_list {
        struct list_head list;
        unsigned long addr;
-       unsigned long vaddr;
        size_t size;
        int type;
 };