From: Paolo Bonzini <pbonzini@redhat.com>
To: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: peter.maydell@linaro.org, Paul Mackerras <paulus@samba.org>,
qemu-devel@nongnu.org, Avi Kivity <avi.kivity@gmail.com>
Subject: Re: [Qemu-devel] [PATCH v2 03/17] memory: add ref/unref calls
Date: Tue, 25 Jun 2013 10:49:39 +0200 [thread overview]
Message-ID: <51C959A3.9030504@redhat.com> (raw)
In-Reply-To: <51BAEBC4.4030607@ozlabs.ru>
Il 14/06/2013 12:09, Alexey Kardashevskiy ha scritto:
> Hi.
>
> Ok. Back to the bug with this patch. The initial problem with this patch is
> that "make check" fails.
>
> Please help with subpages.
>
> It turned out that tests use MALLOC_PERTURB_ which is normally off. Who
> does not know - this is a way to tell glibc to fill released memory with
> some value and then debug accesses to released memory. Some bright mind
> made it random what confuses a lot (and btw valgrind found nothing :-/ ).
> So I spend some time before figured out how to reproduce it outside of the
> qtest thingy.
>
> The tree is qemu.org/master "bd5c51e Michael Roth qemu-char: don't issue
> CHR_EVENT_OPEN in a BH" + replayed patches till the one from $subj on top
> of it. QEMU is configured as "configure --target-list=x86_64-softmmu".
>
> The magic is:
>
> export MALLOC_PERTURB_=123
> nc -l -U ~/qtest-16318.sock &
> nc -l -U ~/qtest-16318.qmp &
> x86_64-softmmu/qemu-system-x86_64 \
> -qtest unix:/home/alexey/qtest-16318.sock,nowait \
> -qtest-log /dev/null \
> -qmp unix:/home/alexey/qtest-16318.qmp,nowait \
> -pidfile ~/qtest-16318.pid -machine accel=qtest -vnc none
>
> Immediate crash at (the very last backtrace in this mail is that crash).
It's a use-after-free, strange that valgrind found nothing. The
subpage is freed in destroy_page_desc before being used in
phys_sections_clear.
The fix is to move freeing the phys_sections in phys_sections_clear,
which is also cleaner.
diff --git a/exec.c b/exec.c
index fa5de88..93907e9 100644
--- a/exec.c
+++ b/exec.c
@@ -762,17 +762,6 @@ hwaddr memory_region_section_get_iotlb(CPUArchState *env,
static int subpage_register (subpage_t *mmio, uint32_t start, uint32_t end,
uint16_t section);
static subpage_t *subpage_init(AddressSpace *as, hwaddr base);
-static void destroy_page_desc(uint16_t section_index)
-{
- MemoryRegionSection *section = &phys_sections[section_index];
- MemoryRegion *mr = section->mr;
-
- if (mr->subpage) {
- subpage_t *subpage = container_of(mr, subpage_t, iomem);
- memory_region_destroy(&subpage->iomem);
- g_free(subpage);
- }
-}
static void destroy_l2_mapping(PhysPageEntry *lp, unsigned level)
{
@@ -787,8 +776,6 @@ static void destroy_l2_mapping(PhysPageEntry *lp, unsigned level)
for (i = 0; i < L2_SIZE; ++i) {
if (!p[i].is_leaf) {
destroy_l2_mapping(&p[i], level - 1);
- } else {
- destroy_page_desc(p[i].ptr);
}
}
lp->is_leaf = 0;
@@ -819,11 +806,20 @@ static uint16_t phys_section_add(MemoryRegionSection *section)
return phys_sections_nb++;
}
+static void phys_section_destroy(MemoryRegion *mr)
+{
+ memory_region_unref(mr);
+
+ if (mr->subpage) {
+ subpage_t *subpage = container_of(mr, subpage_t, iomem);
+ memory_region_destroy(&subpage->iomem);
+ g_free(subpage);
+ }
+}
+
static void phys_sections_clear(void)
{
while (phys_sections_nb > 0) {
MemoryRegionSection *section = &phys_sections[--phys_sections_nb];
- memory_region_unref(section->mr);
+ phys_section_destroy(section->mr);
}
}
next prev parent reply other threads:[~2013-06-25 8:49 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-06-04 12:13 [Qemu-devel] [PATCH v2 00/17] Memory/IOMMU patches part 4: region ownership Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 01/17] memory: add getter/setter for owner Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 02/17] memory: add ref/unref Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 03/17] memory: add ref/unref calls Paolo Bonzini
2013-06-13 6:28 ` Alexey Kardashevskiy
2013-06-13 9:02 ` Alexey Kardashevskiy
2013-06-14 10:09 ` Alexey Kardashevskiy
2013-06-14 13:56 ` Paolo Bonzini
2013-06-14 15:04 ` Alexey Kardashevskiy
2013-06-25 8:49 ` Paolo Bonzini [this message]
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 04/17] exec: add a reference to the region returned by address_space_translate Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 05/17] pci: set owner for BARs Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 06/17] sysbus: add sysbus_pass_mmio Paolo Bonzini
2013-06-04 12:24 ` Peter Maydell
2013-06-04 12:31 ` Paolo Bonzini
2013-06-04 12:36 ` Peter Maydell
2013-06-04 13:24 ` Paolo Bonzini
2013-06-04 14:11 ` Peter Maydell
2013-06-04 14:27 ` Paolo Bonzini
2013-06-04 14:56 ` Peter Maydell
2013-06-04 15:06 ` Paolo Bonzini
2013-06-04 16:50 ` Paolo Bonzini
2013-06-04 17:47 ` Alex Williamson
2013-06-04 19:11 ` Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 07/17] sysbus: set owner for MMIO regions Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 08/17] acpi: add memory_region_set_owner calls Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 09/17] misc: " Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 10/17] isa/portio: allow setting an owner Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 11/17] vga: add memory_region_set_owner calls Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 12/17] pci-assign: " Paolo Bonzini
2013-06-04 16:42 ` Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 13/17] vfio: " Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 14/17] exec: check MRU in qemu_ram_addr_from_host Paolo Bonzini
2013-06-04 12:13 ` [Qemu-devel] [PATCH v2 15/17] exec: move qemu_ram_addr_from_host_nofail to cputlb.c Paolo Bonzini
2013-06-04 12:14 ` [Qemu-devel] [PATCH v2 16/17] memory: return MemoryRegion from qemu_ram_addr_from_host Paolo Bonzini
2013-06-04 12:14 ` [Qemu-devel] [PATCH v2 17/17] memory: ref/unref memory across address_space_map/unmap Paolo Bonzini
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=51C959A3.9030504@redhat.com \
--to=pbonzini@redhat.com \
--cc=aik@ozlabs.ru \
--cc=avi.kivity@gmail.com \
--cc=paulus@samba.org \
--cc=peter.maydell@linaro.org \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).