4:14 a.m.
Hi, Mathias
Thank you for the patchset.
On Tue, Sep 26, 2023 at 10:36 PM <crash-utility-request(a)redhat.com> wrote:
Date: Tue, 26 Sep 2023 12:12:47 +0200
From: Mathias Krause <minipli(a)grsecurity.net>
To: crash-utility(a)redhat.com
Subject: [Crash-utility] [PATCH 2/2] memory_driver: Support overriding
kernel directory
Message-ID: <20230926101247.1237748-3-minipli(a)grsecurity.net>
Content-Type: text/plain; charset="US-ASCII"; x-default=true
Support compiling the module against a different kernel version than the
currently running one by allowing to set either KVER or KDIR variables
on the make commandline.
Also modernize the makefile slightly and make use of the kernel's
'clean' target to ensure to remove all generated files.
The [PATCH 1/2] looks good to me.
For the [PATCH 2/2], I only have two questions:
[1] With the patch 2/2, it always triggers recompiling the gdb like this:
# make lzo
TARGET: PPC64
CRASH: 8.0.3++
GDB: 10.2
CXX gdb.o
CXX ../../crash_target.o
CXX ada-exp.o
CXX ada-lang.o
CXX ada-tasks.o
CXX ada-typeprint.o
CXX ada-valprint.o
CXX ada-varobj.o
CXX addrmap.o
CXX agent.o
CXX alloc.o
CXX annotate.o
...
[2] With the patch 2/2, it always reports the following error "No such file
or directory", if the kernel-devel package is not installed.
# make clean
...
make -C /lib/modules/xxx/build M=/home/crash SUBDIRS=/home/crash clean ||
rm -f *.mod.c *.ko *.o Module.*
make[3]: *** /lib/modules/xxx/build: No such file or directory. Stop.
Actually, I did not build the crash.ko in the directory memory_driver/.
Is that expected behavior?
Thanks
Lianbo
Signed-off-by: Mathias Krause <minipli(a)grsecurity.net>
---
memory_driver/Makefile | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/memory_driver/Makefile b/memory_driver/Makefile
index b494aa3cd184..b720661fa75f 100644
--- a/memory_driver/Makefile
+++ b/memory_driver/Makefile
@@ -8,8 +8,13 @@
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
+ifneq ($(KERNELRELEASE),)
obj-m := crash.o
+else
+KVER ?= $(shell uname -r)
+KDIR ?= /lib/modules/${KVER}/build
all:
- make -C /lib/modules/`uname -r`/build M=${PWD} SUBDIRS=${PWD}
modules
+ ${MAKE} -C ${KDIR} M=${PWD} SUBDIRS=${PWD} modules
clean:
- rm -f *.mod.c *.ko *.o Module.*
+ ${MAKE} -C ${KDIR} M=${PWD} SUBDIRS=${PWD} clean || ${RM} *.mod.c
*.ko *.o Module.*
+endif
--
2.30.2
1:50 p.m.
New subject: [PATCH 2/2] memory_driver: Support overriding kernel directory
Hi Lianbo,
Thanks for testing! My response is inline below...
On 27.09.23 11:14, lijiang wrote:
Hi, Mathias
Thank you for the patchset.
On Tue, Sep 26, 2023 at 10:36 PM <crash-utility-request(a)redhat.com
<mailto:crash-utility-request@redhat.com>> wrote:
Date: Tue, 26 Sep 2023 12:12:47 +0200
From: Mathias Krause <minipli(a)grsecurity.net
<mailto:minipli@grsecurity.net>>
To: crash-utility(a)redhat.com <mailto:crash-utility@redhat.com>
Subject: [Crash-utility] [PATCH 2/2] memory_driver: Support overriding
kernel directory
Message-ID: <20230926101247.1237748-3-minipli(a)grsecurity.net
<mailto:20230926101247.1237748-3-minipli@grsecurity.net>>
Content-Type: text/plain; charset="US-ASCII"; x-default=true
Support compiling the module against a different kernel version than the
currently running one by allowing to set either KVER or KDIR variables
on the make commandline.
Also modernize the makefile slightly and make use of the kernel's
'clean' target to ensure to remove all generated files.
The [PATCH 1/2] looks good to me.
Thanks.
For the [PATCH 2/2], I only have two questions:
[1] With the patch 2/2, it always triggers recompiling the gdb like this:
# make lzo
TARGET: PPC64
CRASH: 8.0.3++
GDB: 10.2
CXX gdb.o
CXX ../../crash_target.o
CXX ada-exp.o
CXX ada-lang.o
CXX ada-tasks.o
CXX ada-typeprint.o
CXX ada-valprint.o
CXX ada-varobj.o
CXX addrmap.o
CXX agent.o
CXX alloc.o
CXX annotate.o
...
Hmm, I cannot reproduce this. If I do 'make lzo' it builds 'crash' only
once, not multiple times as you're observing. Can you please provide
instructions how to reproduce the issue? Or, maybe, it's related to the
second one...
[2] With the patch 2/2, it always reports the following error "No such
file or directory", if the kernel-devel package is not installed.
# make clean
...
make -C /lib/modules/xxx/build M=/home/crash SUBDIRS=/home/crash clean
|| rm -f *.mod.c *.ko *.o Module.*
make[3]: *** /lib/modules/xxx/build: No such file or directory. Stop.
Actually, I did not build the crash.ko in the directory memory_driver/.
Is that expected behavior?
Ahh, yes. That's expected, but unfortunate behavior. I'll fix it by
testing if the build directory exists first to avoid the error message.
The fall-back handling is already there -- simply does the old 'rm
*.mod.c *.ko *.o Module.*'. However, there's no need to generate an
error message when we know in advance that the kernel build directory is
missing.
I'll send v2 in a moment.
Thanks,
Mathias
Thanks
Lianbo
Signed-off-by: Mathias Krause <minipli(a)grsecurity.net
<mailto:minipli@grsecurity.net>>
---
memory_driver/Makefile | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/memory_driver/Makefile b/memory_driver/Makefile
index b494aa3cd184..b720661fa75f 100644
--- a/memory_driver/Makefile
+++ b/memory_driver/Makefile
@@ -8,8 +8,13 @@
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
+ifneq ($(KERNELRELEASE),)
obj-m := crash.o
+else
+KVER ?= $(shell uname -r)
+KDIR ?= /lib/modules/${KVER}/build
all:
- make -C /lib/modules/`uname -r`/build M=${PWD}
SUBDIRS=${PWD} modules
+ ${MAKE} -C ${KDIR} M=${PWD} SUBDIRS=${PWD} modules
clean:
- rm -f *.mod.c *.ko *.o Module.*
+ ${MAKE} -C ${KDIR} M=${PWD} SUBDIRS=${PWD} clean || ${RM}
*.mod.c *.ko *.o Module.*
+endif
--
2.30.2
10:13 p.m.
New subject: [PATCH 2/2] memory_driver: Support overriding kernel directory
On Thu, Sep 28, 2023 at 2:50 AM Mathias Krause <minipli(a)grsecurity.net>
wrote:
Hi Lianbo,
Thanks for testing! My response is inline below...
On 27.09.23 11:14, lijiang wrote:
> Hi, Mathias
> Thank you for the patchset.
> On Tue, Sep 26, 2023 at 10:36 PM <crash-utility-request(a)redhat.com
> <mailto:crash-utility-request@redhat.com>> wrote:
>
> Date: Tue, 26 Sep 2023 12:12:47 +0200
> From: Mathias Krause <minipli(a)grsecurity.net
> <mailto:minipli@grsecurity.net>>
> To: crash-utility(a)redhat.com <mailto:crash-utility@redhat.com>
> Subject: [Crash-utility] [PATCH 2/2] memory_driver: Support
overriding
> kernel directory
> Message-ID: <20230926101247.1237748-3-minipli(a)grsecurity.net
> <mailto:20230926101247.1237748-3-minipli@grsecurity.net>>
> Content-Type: text/plain; charset="US-ASCII"; x-default=true
>
> Support compiling the module against a different kernel version than
the
> currently running one by allowing to set either KVER or KDIR
variables
> on the make commandline.
>
> Also modernize the makefile slightly and make use of the kernel's
> 'clean' target to ensure to remove all generated files.
>
>
> The [PATCH 1/2] looks good to me.
Thanks.
>
> For the [PATCH 2/2], I only have two questions:
>
> [1] With the patch 2/2, it always triggers recompiling the gdb like this:
> # make lzo
> TARGET: PPC64
> CRASH: 8.0.3++
> GDB: 10.2
>
> CXX gdb.o
> CXX ../../crash_target.o
> CXX ada-exp.o
> CXX ada-lang.o
> CXX ada-tasks.o
> CXX ada-typeprint.o
> CXX ada-valprint.o
> CXX ada-varobj.o
> CXX addrmap.o
> CXX agent.o
> CXX alloc.o
> CXX annotate.o
> ...
Hmm, I cannot reproduce this. If I do 'make lzo' it builds 'crash' only
once, not multiple times as you're observing. Can you please provide
instructions how to reproduce the issue? Or, maybe, it's related to the
second one...
This issue still exists with the v2:
https://listman.redhat.com/archives/crash-utility/2023-September/011021.html
I can reproduce it with the following steps(require kernel-devel package):
[1] make lzo
[2] make clean
[3] make lzo(or make warn)
# make lzo
TARGET: ARM64
CRASH: 8.0.3++
GDB: 10.2
CXX gdb.o
CXX ../../crash_target.o
CXX aarch32-linux-nat.o
CXX aarch32-tdep.o
CXX aarch64-linux-nat.o
CXX aarch64-linux-tdep.o
CXX aarch64-ravenscar-thread.o
CXX aarch64-tdep.o
...
# make warn
TARGET: ARM64
CRASH: 8.0.3++
GDB: 10.2
CXX gdb.o
CXX ../../crash_target.o
CXX aarch32-linux-nat.o
CXX aarch32-tdep.o
CXX aarch64-linux-nat.o
CXX aarch64-linux-tdep.o
CXX aarch64-ravenscar-thread.o
CXX aarch64-tdep.o
...
> [2] With the patch 2/2, it always reports the following error "No such
> file or directory", if the kernel-devel package is not installed.
> # make clean
> ...
> make -C /lib/modules/xxx/build M=/home/crash SUBDIRS=/home/crash clean
> || rm -f *.mod.c *.ko *.o Module.*
> make[3]: *** /lib/modules/xxx/build: No such file or directory. Stop.
>
> Actually, I did not build the crash.ko in the directory memory_driver/.
>
> Is that expected behavior?
Ahh, yes. That's expected, but unfortunate behavior. I'll fix it by
testing if the build directory exists first to avoid the error message.
The fall-back handling is already there -- simply does the old 'rm
*.mod.c *.ko *.o Module.*'. However, there's no need to generate an
error message when we know in advance that the kernel build directory is
missing.
With the v2, this issue disappeared.
Thanks.
Lianbo
3:50 a.m.
New subject: [PATCH 2/2] memory_driver: Support overriding kernel directory
On 28.09.23 05:13, lijiang wrote:
[...]
This issue still exists with the v2:
https://listman.redhat.com/archives/crash-utility/2023-September/011021.html
<https://listman.redhat.com/archives/crash-utility/2023-September/011021.h...
I can reproduce it with the following steps(require kernel-devel package):
[1] make lzo
[2] make clean
[3] make lzo(or make warn)
# make lzo
TARGET: ARM64
CRASH: 8.0.3++
GDB: 10.2
CXX gdb.o
CXX ../../crash_target.o
CXX aarch32-linux-nat.o
CXX aarch32-tdep.o
CXX aarch64-linux-nat.o
CXX aarch64-linux-tdep.o
CXX aarch64-ravenscar-thread.o
CXX aarch64-tdep.o
...
# make warn
TARGET: ARM64
CRASH: 8.0.3++
GDB: 10.2
CXX gdb.o
CXX ../../crash_target.o
CXX aarch32-linux-nat.o
CXX aarch32-tdep.o
CXX aarch64-linux-nat.o
CXX aarch64-linux-tdep.o
CXX aarch64-ravenscar-thread.o
CXX aarch64-tdep.o
...
Thanks, I can see build problems now too after rebasing my changes to
the current master branch.
Digging into, I noticed that building the memory_driver is actually
broken, as it passes the wrong path to the kernel's makefile.
When invoked via '${MAKE} -C memory_driver ...' $(PWD) still refers to
the upper directory, i.e. .../crash and not .../crash/memory_driver.
That's why the '${MAKE} -C ${KDIR} M=${PWD} ... clean' actually calls
the 'clean' target in .../crash and not .../crash/memory_driver as it
should.
The fix is easy, $(PWD) needs to be updated in memory_driver/Makefile.
I'll send a v3 with this change broken out as a separate patch, as it is
a bugfix on its own ('make memory_driver' does nothing in master right now).
>
> [2] With the patch 2/2, it always reports the following error "No such
> file or directory", if the kernel-devel package is not installed.
> # make clean
> ...
> make -C /lib/modules/xxx/build M=/home/crash SUBDIRS=/home/crash clean
> || rm -f *.mod.c *.ko *.o Module.*
> make[3]: *** /lib/modules/xxx/build: No such file or directory. Stop.
>
> Actually, I did not build the crash.ko in the directory
memory_driver/.
>
> Is that expected behavior?
Ahh, yes. That's expected, but unfortunate behavior. I'll fix it by
testing if the build directory exists first to avoid the error message.
The fall-back handling is already there -- simply does the old 'rm
*.mod.c *.ko *.o Module.*'. However, there's no need to generate an
error message when we know in advance that the kernel build directory is
missing.
With the v2, this issue disappeared.
Thanks,
Mathias
Thanks.
Lianbo
425
days inactive
426
days old
devel@lists.crash-utility.osci.io
3 comments
2 participants
participants (2)
-
lijiang -
Mathias Krause