Re: [PATCH] kexec: Enable runtime allocation of crash_image
From: Philipp Rudo
Date: Tue Nov 29 2022 - 10:14:01 EST
Ricardo,
On Mon, 28 Nov 2022 18:07:06 +0100
Ricardo Ribalda <ribalda@xxxxxxxxxxxx> wrote:
> Hi Philipp
>
>
> Thanks for your review.
>
>
> On Mon, 28 Nov 2022 at 18:00, Philipp Rudo <prudo@xxxxxxxxxx> wrote:
> >
> > Hi Ricardo,
> >
> > On Thu, 24 Nov 2022 23:23:36 +0100
> > Ricardo Ribalda <ribalda@xxxxxxxxxxxx> wrote:
> >
> > > Usually crash_image is defined statically via the crashkernel parameter
> > > or DT.
> > >
> > > But if the crash kernel is not used, or is smaller than then
> > > area pre-allocated that memory is wasted.
> > >
> > > Also, if the crash kernel was not defined at bootime, there is no way to
> > > use the crash kernel.
> > >
> > > Enable runtime allocation of the crash_image if the crash_image is not
> > > defined statically. Following the same memory allocation/validation path
> > > that for the reboot kexec kernel.
> > >
> > > Signed-off-by: Ricardo Ribalda <ribalda@xxxxxxxxxxxx>
> >
> > I don't think this patch will work as intended. For one you omit
> > setting the image->type to KEXEC_TYPE_CRASH. But when you grep for that
> > type you will find that there is a lot of special handling done for it.
> > I don't believe that this can simply be skipped without causing
> > problems.
> >
> > Furthermore I think you have missed one important detail. The memory
> > reserved for the crash kernel is not just a buffer for the image but
> > the memory it runs in! For that it has to be a continuous piece of
> > physical memory with usually some additional arch specific limitations.
> > When allocated dynamically all those limitations need to be considered.
> > But a standard kexec doesn't care about those limitations as it doesn't
> > care about the os running before itself. It can simply overwrite the
> > memory when booting. But if the crash kernel does the same it will
> > corrupt the dump it is supposed to generate.
>
> Right now, I do not intend to use it to fetch a kdump, I am using it
> as the image that will run when the system crashes.
the crash_image is currently all about creating a dump. If you want to
change that you need to discuss the new behavior in the commit message!
Please update the commit message.
Thanks
Philipp
>
> It seems to work fine on the two devices that I am using for tests.
>
> >
> > Thanks
> > Philipp
> >
> > > ---
> > > kexec: Enable runtime allocation of crash_image
> > >
> > > To: Eric Biederman <ebiederm@xxxxxxxxxxxx>
> > > Cc: kexec@xxxxxxxxxxxxxxxxxxx
> > > Cc: linux-kernel@xxxxxxxxxxxxxxx
> > > Cc: Sergey Senozhatsky <senozhatsky@xxxxxxxxxxxx>
> > > Cc: linux-kernel@xxxxxxxxxxxxxxx
> > > Cc: Ross Zwisler <zwisler@xxxxxxxxxx>
> > > Cc: Philipp Rudo <prudo@xxxxxxxxxx>
> > > Cc: Baoquan He <bhe@xxxxxxxxxx>
> > > ---
> > > include/linux/kexec.h | 1 +
> > > kernel/kexec.c | 9 +++++----
> > > kernel/kexec_core.c | 5 +++++
> > > kernel/kexec_file.c | 7 ++++---
> > > 4 files changed, 15 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/include/linux/kexec.h b/include/linux/kexec.h
> > > index 41a686996aaa..98ca9a32bc8e 100644
> > > --- a/include/linux/kexec.h
> > > +++ b/include/linux/kexec.h
> > > @@ -427,6 +427,7 @@ extern int kexec_load_disabled;
> > > extern bool kexec_in_progress;
> > >
> > > int crash_shrink_memory(unsigned long new_size);
> > > +bool __crash_memory_valid(void);
> > > ssize_t crash_get_memory_size(void);
> > >
> > > #ifndef arch_kexec_protect_crashkres
> > > diff --git a/kernel/kexec.c b/kernel/kexec.c
> > > index cb8e6e6f983c..b5c17db25e88 100644
> > > --- a/kernel/kexec.c
> > > +++ b/kernel/kexec.c
> > > @@ -28,7 +28,7 @@ static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
> > > struct kimage *image;
> > > bool kexec_on_panic = flags & KEXEC_ON_CRASH;
> > >
> > > - if (kexec_on_panic) {
> > > + if (kexec_on_panic && __crash_memory_valid()) {
> > > /* Verify we have a valid entry point */
> > > if ((entry < phys_to_boot_phys(crashk_res.start)) ||
> > > (entry > phys_to_boot_phys(crashk_res.end)))
> > > @@ -44,7 +44,7 @@ static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
> > > image->nr_segments = nr_segments;
> > > memcpy(image->segment, segments, nr_segments * sizeof(*segments));
> > >
> > > - if (kexec_on_panic) {
> > > + if (kexec_on_panic && __crash_memory_valid()) {
> > > /* Enable special crash kernel control page alloc policy. */
> > > image->control_page = crashk_res.start;
> > > image->type = KEXEC_TYPE_CRASH;
> > > @@ -101,7 +101,7 @@ static int do_kexec_load(unsigned long entry, unsigned long nr_segments,
> > >
> > > if (flags & KEXEC_ON_CRASH) {
> > > dest_image = &kexec_crash_image;
> > > - if (kexec_crash_image)
> > > + if (kexec_crash_image && __crash_memory_valid())
> > > arch_kexec_unprotect_crashkres();
> > > } else {
> > > dest_image = &kexec_image;
> > > @@ -157,7 +157,8 @@ static int do_kexec_load(unsigned long entry, unsigned long nr_segments,
> > > image = xchg(dest_image, image);
> > >
> > > out:
> > > - if ((flags & KEXEC_ON_CRASH) && kexec_crash_image)
> > > + if ((flags & KEXEC_ON_CRASH) && kexec_crash_image &&
> > > + __crash_memory_valid())
> > > arch_kexec_protect_crashkres();
> > >
> > > kimage_free(image);
> > > diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c
> > > index ca2743f9c634..77083c9760fb 100644
> > > --- a/kernel/kexec_core.c
> > > +++ b/kernel/kexec_core.c
> > > @@ -1004,6 +1004,11 @@ void crash_kexec(struct pt_regs *regs)
> > > }
> > > }
> > >
> > > +bool __crash_memory_valid(void)
> > > +{
> > > + return crashk_res.end != crashk_res.start;
> > > +}
> > > +
> > > ssize_t crash_get_memory_size(void)
> > > {
> > > ssize_t size = 0;
> > > diff --git a/kernel/kexec_file.c b/kernel/kexec_file.c
> > > index 45637511e0de..0671f4f370ff 100644
> > > --- a/kernel/kexec_file.c
> > > +++ b/kernel/kexec_file.c
> > > @@ -280,7 +280,7 @@ kimage_file_alloc_init(struct kimage **rimage, int kernel_fd,
> > >
> > > image->file_mode = 1;
> > >
> > > - if (kexec_on_panic) {
> > > + if (kexec_on_panic && __crash_memory_valid()) {
> > > /* Enable special crash kernel control page alloc policy. */
> > > image->control_page = crashk_res.start;
> > > image->type = KEXEC_TYPE_CRASH;
> > > @@ -345,7 +345,7 @@ SYSCALL_DEFINE5(kexec_file_load, int, kernel_fd, int, initrd_fd,
> > > dest_image = &kexec_image;
> > > if (flags & KEXEC_FILE_ON_CRASH) {
> > > dest_image = &kexec_crash_image;
> > > - if (kexec_crash_image)
> > > + if (kexec_crash_image && __crash_memory_valid())
> > > arch_kexec_unprotect_crashkres();
> > > }
> > >
> > > @@ -408,7 +408,8 @@ SYSCALL_DEFINE5(kexec_file_load, int, kernel_fd, int, initrd_fd,
> > > exchange:
> > > image = xchg(dest_image, image);
> > > out:
> > > - if ((flags & KEXEC_FILE_ON_CRASH) && kexec_crash_image)
> > > + if ((flags & KEXEC_FILE_ON_CRASH) && kexec_crash_image &&
> > > + __crash_memory_valid())
> > > arch_kexec_protect_crashkres();
> > >
> > > kexec_unlock();
> > >
> > > ---
> > > base-commit: 4312098baf37ee17a8350725e6e0d0e8590252d4
> > > change-id: 20221124-kexec-noalloc-3cab3cbe000f
> > >
> > > Best regards,
> >
>
>