"basic" system tests fail (and all the other ones) on guix master

Done

Details

7 participants

Danny Milosavljevic
Oleg Pykhalov
Leo Famulari
Ludovic Courtès
Marius Bakke
Mark H Weaver
Mathieu Othacehe

Owner: unassigned

Submitted by: Danny Milosavljevic

Severity: normal

Merged with

43352

Debbugs page

Danny Milosavljevic wrote 4 years ago

Recipients:(address . bug-guix@gnu.org)

Message-ID:20200911195058.6dc013b4@scratchpost.org

Hi,

as of guix master commit 0fb974be9c3e1e22a2145c9c602c44cd10cef2b0 all system
tests, including "basic", fail:

$ guix environment --pure guix --ad-hoc git guile-readline guile-json nano guile-zlib guile-lzlib
(env)$ make TESTS=basic check-system
loading '/gnu/store/s3limrgxj4pd6b4psra66phary2nmqx4-linux-vm-loader'...
[    2.776050] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
environment variable `PATH' set to `/gnu/store/j3jlpncfqvykkq6sx7h4ly1rdcr2a8qq'
creating partition table with 2 partitions (20.0 MiB, 40.0 MiB)...
Error: Partition(s) 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, .
Backtrace:
           3 (primitive-load "/gnu/store/liacrhaiab35pcfzdqsa71fvp7f?")
In ./gnu/build/vm.scm:
   470:21  2 (initialize-hard-disk "/dev/vda" #:bootloader-package _ ?)
    300:2  1 (initialize-partition-table "/dev/vda" (#<<partitio???@ build-log 1
?@ build-log 29023 1
?@ build-log 29023 1

Toggle quote (1 lines)

>@ build-log 29023 1

 @ build-log 29023 1
?@ build-log 29023 1
?@ build-log 29023 1
?@ build-log 29023 1
)@ build-log 29023 1
 @ build-log 29023 1
?@ build-log 29023 1
?@ build-log 29023 1
?@ build-log 29023 1
)@ build-log 29023 1
@ build-log 29023 1

In ./guix/build/utils.scm:
    654:6  0 (invoke _ . _)

./guix/build/utils.scm:654:6: In procedure invoke:
ERROR:
  1. &invoke-error:
      program: "parted"
      arguments: ("--script" "/dev/vda" "mklabel" "msdos" "mkpart" "primary" "e)
      exit-status: 1
      term-signal: #f
      stop-signal: #f
[    3.446031] Unregister pv shared memory for cpu 0
[    3.447989] reboot: Restarting system
[    3.449167] reboot: machine restart
Backtrace: 
           1 (primitive-load "/gnu/store/1xp94hddh88szwkbqyb49jirmf5?")
In ./gnu/build/vm.scm:
   198:12  0 (load-in-linux-vm _ #:output _ #:qemu _ #:memory-size _ ?)

./gnu/build/vm.scm:198:12: In procedure load-in-linux-vm:
guest VM code exited with a non-zero status 256
note: keeping build directory `/tmp/guix-build-qemu-image.drv-0'
builder for `/gnu/store/ks5w1bwcm2rg1fq5s9hc78grxpnm0wpf-qemu-image.drv' failed1
build of /gnu/store/ks5w1bwcm2rg1fq5s9hc78grxpnm0wpf-qemu-image.drv failed
View build log at '/var/log/guix/drvs/ks/5w1bwcm2rg1fq5s9hc78grxpnm0wpf-qemu-im.
cannot build derivation `/gnu/store/8vhwchcxlns64nnqr13p0liw745s6pp8-run-vm.sh.t
cannot build derivation `/gnu/store/672vqgm8mx0a6k2iwsf13q73gmk5vdc4-basic.drv't
guix build: error: build of `/gnu/store/672vqgm8mx0a6k2iwsf13q73gmk5vdc4-basic.d
make: *** [Makefile:6007: check-system] Error 1
dannym@bayfront ~/src/guix-master/guix [env]$

The same happens on my laptop.

./pre-inst-env guix describe
Git checkout:
  repository: /home/dannym/src/guix-master/guix
  branch: master
  commit: 0fb974be9c3e1e22a2145c9c602c44cd10cef2b0

(ssh://dannym@git.sv.gnu.org/srv/git/guix.git)

I've tried to add a call to wipefs before parted--it doesn't help.

Before the problem surfaced I saw a qemu major update.

-----BEGIN PGP SIGNATURE-----

iQEzBAEBCgAdFiEEds7GsXJ0tGXALbPZ5xo1VCwwuqUFAl9buQIACgkQ5xo1VCww

uqWBhAf9FvsI4m5s/5ARUyX7W0wwU9CqTVw9eGwTXcNlD+8GvOBTTDYrO0ERbcPj

3hEioTvIGlZJ33ic02ZTrewMWjmlbRewrB3E3vTpFNB21gj9OYP5ptm0amPKOtM9

kvUzxNUO7jpFspUv/dmFvW1d7zxbwGjXyNFQbKgik5P57QRxqhenEORGOwHdlPDD

1iHwN9ZAOQoT0neRrCWChPs3wWxXFzfaQGfTlcC3hGCZ+pRgrvEo0OsvkINc85t4

h65WB8vO8OlYtQA9BgRYMM58uUk9tZ44U+6+PFZ1DcUHfeac6/8LTdSixCSXW2uj

h3Viu0E+twhaxqMGqtXqPDRCGsw1ew==

=SYEI

-----END PGP SIGNATURE-----

Oleg Pykhalov wrote 4 years ago

merge 43344 43352

Recipients:(name . control)(address . control@debbugs.gnu.org)

Message-ID:878sdfz2ic.fsf@gmail.com

merge 43344 43352

Oleg Pykhalov wrote 4 years ago

control message for bug #43344

Recipients:(address . control@debbugs.gnu.org)

Message-ID:87y2lfgumt.fsf@gmail.com

merge 43344 43352

quit

Mathieu Othacehe wrote 4 years ago

Re: bug#43344: "basic" system tests fail (and all the other ones) on guix master

Recipients:(name . Danny Milosavljevic)(address . dannym@scratchpost.org)(address . 43344@debbugs.gnu.org)

Message-ID:877dsw7alf.fsf@gnu.org

Hello Danny,

Toggle quote (4 lines)

> 1. &invoke-error:

> program: "parted"

> arguments: ("--script" "/dev/vda" "mklabel" "msdos" "mkpart" "primary" "e)

So it looks like the parted script failed inside the VM, while running

"initialize-partition-table".

This work fine both on the build farm and on my machine, which makes the

debug harder :(.

Anything special with your hardware? KVM support disabled maybe?

Thanks,

Mathieu

Danny Milosavljevic wrote 4 years ago

Recipients:(name . Mathieu Othacehe)(address . othacehe@gnu.org)(address . 43344@debbugs.gnu.org)

Message-ID:20200914183037.624cc347@scratchpost.org

Hi Mathieu,

On Mon, 14 Sep 2020 15:26:52 +0200

Mathieu Othacehe <othacehe@gnu.org> wrote:

Toggle quote (2 lines)

> Anything special with your hardware? KVM support disabled maybe?

The culprit had been the Linux kernel update to 5.8.8.

After downgrading to 5.8.7 it works just fine--no other changes done.

Previously, I had tried also to add wipefs -a before the parted--that

hadn't helped either.

-----BEGIN PGP SIGNATURE-----

iQEzBAEBCgAdFiEEds7GsXJ0tGXALbPZ5xo1VCwwuqUFAl9fms4ACgkQ5xo1VCww

uqWPnwf+M7BFkY+l76yRBwo3WFAVJeGw659xThs3XgEDPMx3slTamBRBbrglOk/n

djxolbFKNwdLV+0KjUoctZ1/9BHKn7vFNABbzzoXc/F86S3O57mqZeKLgHN8g6MY

UJbDVmTsVqcj7bTUf5H4hN0FT0Wl5jByAP7nnpy3jFyEF3Gc9DQshT/I9WoA5lt5

srTlXFncHrdCZ919CkTiRk8ztwVJA5xGIjAZ7D6rEnixtMe76XkwvWzBIgaC0OL9

V7+pmngloNfzunl/msfXuzPPNkU5cnEWK2PIHeyPuACoexu8LysS2Tupf3PzNBDZ

RmUUNsERytsROsAqJHHoAo87ms0UFw==

=AsAG

-----END PGP SIGNATURE-----

Mark H Weaver wrote 4 years ago

Recipients:

Message-ID:87y2laactz.fsf@netris.org

Hi,

Danny Milosavljevic <dannym@scratchpost.org> writes:

Toggle quote (12 lines)> On Mon, 14 Sep 2020 15:26:52 +0200
> Mathieu Othacehe <othacehe@gnu.org> wrote:
>
>> Anything special with your hardware? KVM support disabled maybe?
>
> The culprit had been the Linux kernel update to 5.8.8.
>
> After downgrading to 5.8.7 it works just fine--no other changes done.
>
> Previously, I had tried also to add wipefs -a before the parted--that
> hadn't helped either.

That's useful information, but we should not stay frozen on the 5.8.7

kernel for much longer. 5.8.8 contains many bug fixes, some of which

might fix potentially exploitable flaws.

It would be useful to know if this problem still occurs with 5.8.9,

which has since come out. If so, we should do a bisection between 5.8.7

and 5.8.8 to find out which upstream commit introduced the problem.

Would anyone like to investigate this further?

Mark

Leo Famulari wrote 4 years ago

Recipients:(name . Mark H Weaver)(address . mhw@netris.org)

Message-ID:20200915230628.GA20807@jasmine.lan

On Tue, Sep 15, 2020 at 06:34:21PM -0400, Mark H Weaver wrote:

Toggle quote (10 lines)> That's useful information, but we should not stay frozen on the 5.8.7
> kernel for much longer.  5.8.8 contains many bug fixes, some of which
> might fix potentially exploitable flaws.
> 
> It would be useful to know if this problem still occurs with 5.8.9,
> which has since come out.  If so, we should do a bisection between 5.8.7
> and 5.8.8 to find out which upstream commit introduced the problem.
> 
> Would anyone like to investigate this further?

I will try to reproduce the bug with 5.8.9 now. I will try the bisection

if time permits.

Ludovic Courtès wrote 4 years ago

Recipients:(name . Danny Milosavljevic)(address . dannym@scratchpost.org)

Message-ID:87pn6ln7q3.fsf@gnu.org

Hi,

Danny Milosavljevic <dannym@scratchpost.org> skribis:

Toggle quote (4 lines)

> environment variable `PATH' set to `/gnu/store/j3jlpncfqvykkq6sx7h4ly1rdcr2a8qq'

> creating partition table with 2 partitions (20.0 MiB, 40.0 MiB)...

> Error: Partition(s) 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, .

[...]

Toggle quote (7 lines)

> 1. &invoke-error:

> program: "parted"

> arguments: ("--script" "/dev/vda" "mklabel" "msdos" "mkpart" "primary" "e)

> exit-status: 1

> term-signal: #f

> stop-signal: #f

The code in question in Parted:

Toggle snippet (38 lines)                if (!add_partition (disk, part)) {
                        ok[i - 1] = 0;
                        errnums[i - 1] = errno;
                }

[…]

        char *bad_part_list = NULL;
        /* now warn about any errors */
        for (i = 1; i <= lpn; i++) {
                if (ok[i - 1] || errnums[i - 1] == ENXIO)
                        continue;
                if (bad_part_list == NULL) {
                        bad_part_list = malloc (lpn * 5);
                        if (!bad_part_list)
                                goto cleanup;
                        bad_part_list[0] = 0;
                }
                sprintf (bad_part_list + strlen (bad_part_list), "%d, ", i);
        }
        if (bad_part_list == NULL)
                ret = 1;
        else {
                bad_part_list[strlen (bad_part_list) - 2] = 0;
                if (ped_exception_throw (
                        PED_EXCEPTION_ERROR,
                        PED_EXCEPTION_IGNORE_CANCEL,
                        _("Partition(s) %s on %s have been written, but we have "
                          "been unable to inform the kernel of the change, "
                          "probably because it/they are in use.  As a result, "
                          "the old partition(s) will remain in use.  You "
                          "should reboot now before making further changes."),
                        bad_part_list, disk->dev->path) == PED_EXCEPTION_IGNORE)
                        ret = 1;
                free (bad_part_list);
        }

With the patch below, I strace’d ‘parted’, which gives:

Toggle snippet (15 lines)

$ make check-system TESTS=basic

[…]

ioctl(3, BLKPG, {op=BLKPG_DEL_PARTITION, flags=0, datalen=152, data={start=0, length=0, pno=253, devname="", volname=""}}) = -1 ENOMEM (Cannot allocate memory)

ioctl(3, BLKPG, {op=BLKPG_DEL_PARTITION, flags=0, datalen=152, data={start=0, length=0, pno=254, devname="", volname=""}}) = -1 ENOMEM (Cannot allocate memory)

ioctl(3, BLKPG, {op=BLKPG_DEL_PARTITION, flags=0, datalen=152, data={start=0, length=0, pno=255, devname="", volname=""}}) = -1 ENOMEM (Cannot allocate memory)

ioctl(3, BLKPG, {op=BLKPG_DEL_PARTITION, flags=0, datalen=152, data={start=0, length=0, pno=256, devname="", volname=""}}) = -1 ENOMEM (Cannot allocate memory)

write(2, "Error", 5Error) = 5

write(2, ": ", 2: ) = 2

write(2, "Partition(s) 1, 2, 3, 4, 5, 6, 7"..., 495Partition(s) 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64 on /dev/vda have been written, but we have been unable to inform the kernel of the change, probably because it/they are in use. As a result, the old partition(s) will remain in use. You should reboot now before making further changes.) = 495

write(2, "\n", 1

) = 1

So I threw more virtual RAM at it:

Toggle diff (9 lines)diff --git a/gnu/system/vm.scm b/gnu/system/vm.scm
@@ -446,6 +450,7 @@ system that is passed to 'populate-root-file-system'."
                                      #:bootloader-installer
                                      #+(bootloader-installer bootloader)))))))
    #:system system
+   #:memory-size 1024
    #:make-disk-image? #t
    #:disk-image-size disk-image-size
    #:disk-image-format disk-image-format

… but that doesn’t help.

Ideas?

Thanks,

Ludo’.

Toggle diff (51 lines)diff --git a/gnu/build/vm.scm b/gnu/build/vm.scm
index 287d099f79..e793b5b518 100644
--- a/gnu/build/vm.scm
+++ b/gnu/build/vm.scm
@@ -297,7 +297,7 @@ actual /dev name based on DEVICE."
                                      partition-size)
                             partitions)
                        ", "))
-  (apply invoke "parted" "--script"
+  (apply invoke "strace" "parted" "--script"
          device "mklabel" label-type
          (options partitions offset))
 
diff --git a/gnu/packages/linux.scm b/gnu/packages/linux.scm
index 59ffb334e0..72fb3ca49d 100644
--- a/gnu/packages/linux.scm
+++ b/gnu/packages/linux.scm
@@ -349,15 +349,15 @@ corresponding UPSTREAM-SOURCE (an origin), using the given DEBLOB-SCRIPTS."
 
 ;; The current "stable" kernel. That is, the most recently released major
 ;; version.
-(define-public linux-libre-5.8-version "5.8.7")
+(define-public linux-libre-5.8-version "5.8.8")
 (define deblob-scripts-5.8
   (linux-libre-deblob-scripts
-   linux-libre-5.8-version
+   "5.8.7"
    (base32 "07z7sglyrfh0706icqqf3shadf638pvyid9386r661ds5lbsa2mw")
    (base32 "0j6jba5fcddqlb42f95gjl78jisfla4nswqila074gglcrbnl9q7")))
 (define-public linux-libre-5.8-pristine-source
   (let ((version linux-libre-5.8-version)
-        (hash (base32 "1zhpzlhl2ykna2nc70m72wlgyv1pkvkpfssb4k8p5pwlkh1ga2vv")))
+        (hash (base32 "0xm901zvvrwsb9k88la6pb65nybi43bygiyz1z68njwsx6ripxik")))
    (make-linux-libre-source version
                             (%upstream-linux-source version hash)
                             deblob-scripts-5.8)))
diff --git a/gnu/system/vm.scm b/gnu/system/vm.scm
index 80a8618729..49489b6159 100644
--- a/gnu/system/vm.scm
+++ b/gnu/system/vm.scm
@@ -376,6 +376,10 @@ system that is passed to 'populate-root-file-system'."
 
              (set-path-environment-variable "PATH" '("bin" "sbin") inputs)
 
+             (setenv "PATH"
+                     (string-append #+(file-append strace "/bin") ":"
+                                    (getenv "PATH")))
+
              (let* ((graphs     '#$(match inputs
                                      (((names . _) ...)
                                       names)))

Leo Famulari wrote 4 years ago

Recipients:(name . Mark H Weaver)(address . mhw@netris.org)

Message-ID:20200916141231.GA28015@jasmine.lan

On Tue, Sep 15, 2020 at 07:06:28PM -0400, Leo Famulari wrote:

Toggle quote (3 lines)

> I will try to reproduce the bug with 5.8.9 now. I will try the bisection

> if time permits.

It also fails with 5.8.9.

Danny Milosavljevic wrote 4 years ago

Recipients:(name . Leo Famulari)(address . leo@famulari.name)

Message-ID:20200916165245.421e3c4c@scratchpost.org

commit 692d0626557451c4b557397f20b7394b612d0289
Author: Christoph Hellwig <hch@lst.de>
Date:   Tue Sep 1 11:59:41 2020 +0200

    block: fix locking in bdev_del_partition
    
    [ Upstream commit 08fc1ab6d748ab1a690fd483f41e2938984ce353 ]
    
    We need to hold the whole device bd_mutex to protect against
    other thread concurrently deleting out partition before we get
    to it, and thus causing a use after free.
    
    Fixes: cddae808aeb7 ("block: pass a hd_struct to delete_partition")
    Reported-by: syzbot+6448f3c229bc52b82f69@syzkaller.appspotmail.com
    Signed-off-by: Christoph Hellwig <hch@lst.de>
    Signed-off-by: Jens Axboe <axboe@kernel.dk>
    Signed-off-by: Sasha Levin <sashal@kernel.org>

?

-----BEGIN PGP SIGNATURE-----

iQEzBAEBCgAdFiEEds7GsXJ0tGXALbPZ5xo1VCwwuqUFAl9iJr4ACgkQ5xo1VCww

uqVrNwf/dfRyuoleqM4FawaPaWmdJPsd8tDj+OGRGRvSnncbqF1IodkEw6XhMx3y

oDs3fy8irWSDPwkFqiL/pFEd15IfbC4mFg2kTUo1bD/R7euCjZMMvwpIWWUnQ+KY

saA/frug/F9yZRUDRQPWmrJJ/eXX2n6ina1rsaRTYt/yOUgmhpR/wPhme/QZdJaR

CsSnosOiR+qqDrHVQJRlP0nu3bMUNXB3j9D1ylSjRfyb6uuAQWuuNoN/u/jedldJ

2STJRZRITq/T+678CGO4f8yMHcuWFzYFcWR1Cf5JYVDUxz86lBTQi+zDf+ULS6lE

Y/mvBSD0GONr20Mj2tsH3qRBI6Ri0A==

=oQi4

-----END PGP SIGNATURE-----

Danny Milosavljevic wrote 4 years ago

Recipients:(name . Leo Famulari)(address . leo@famulari.name)

Message-ID:20200916170239.580ac59a@scratchpost.org

On Wed, 16 Sep 2020 16:52:45 +0200

Danny Milosavljevic <dannym@scratchpost.org> wrote:

Toggle quote (19 lines)> commit 692d0626557451c4b557397f20b7394b612d0289
> Author: Christoph Hellwig <hch@lst.de>
> Date:   Tue Sep 1 11:59:41 2020 +0200
> 
>     block: fix locking in bdev_del_partition
>     
>     [ Upstream commit 08fc1ab6d748ab1a690fd483f41e2938984ce353 ]
>     
>     We need to hold the whole device bd_mutex to protect against
>     other thread concurrently deleting out partition before we get
>     to it, and thus causing a use after free.
>     
>     Fixes: cddae808aeb7 ("block: pass a hd_struct to delete_partition")
>     Reported-by: syzbot+6448f3c229bc52b82f69@syzkaller.appspotmail.com
>     Signed-off-by: Christoph Hellwig <hch@lst.de>
>     Signed-off-by: Jens Axboe <axboe@kernel.dk>
>     Signed-off-by: Sasha Levin <sashal@kernel.org>
> 

int bdev_del_partition(struct block_device *bdev, int partno)

{

struct block_device *bdevp;

struct hd_struct *part = NULL;

int ret;

bdevp = bdget_disk(bdev->bd_disk, partno);

if (!bdevp)

return -ENOMEM; <--------------

...

}

struct block_device *bdget_disk(struct gendisk *disk, int partno)

{

struct hd_struct *part;

struct block_device *bdev = NULL;

part = disk_get_part(disk, partno);

if (part)

bdev = bdget(part_devt(part));

disk_put_part(part);

return bdev;

}

struct block_device *bdget(dev_t dev)

{

struct block_device *bdev;

struct inode *inode;

inode = iget5_locked(blockdev_superblock, hash(dev),

bdev_test, bdev_set, &dev);

if (!inode)

return NULL; <--------------------

[...]

}

-----BEGIN PGP SIGNATURE-----

iQEzBAEBCgAdFiEEds7GsXJ0tGXALbPZ5xo1VCwwuqUFAl9iKQ8ACgkQ5xo1VCww

uqWWPQf+KN0J6gI0j5qCLy9glxGYn52k+m4TbXE83/Zb68h80/ytGW0F8GIPXCnw

+qSuSEErK+Nr/RBTViAKDAS0X7jcYpMlmL42Exyt3vyEnIYwxN8k3pTwqrYwzMjh

Vr2Srdh3SS/NW7rVcR0dOXsgitEPTUVsvgBjvGaofKvoU0Ny2SYi44nnPKGYqskv

wZE42iRH4c8qZz8+LkjO/MFqjgbOcOFcodrGE9DUjxx4RnpmijHIsoTPKJiZ4qjZ

BrAAN+xj2dBh03dK+ZnCodWLqBZmuz4O/6DsiNkiFxLhSQo+RQY1NaEahz/Gns9s

wti5qCbPep8FkTDzKutOSkq9oUceRg==

=bTO1

-----END PGP SIGNATURE-----

Leo Famulari wrote 4 years ago

Recipients:(name . Danny Milosavljevic)(address . dannym@scratchpost.org)

Message-ID:20200916162244.GA3262@jasmine.lan

On Wed, Sep 16, 2020 at 04:52:45PM +0200, Danny Milosavljevic wrote:

Toggle quote (20 lines)> commit 692d0626557451c4b557397f20b7394b612d0289
> Author: Christoph Hellwig <hch@lst.de>
> Date:   Tue Sep 1 11:59:41 2020 +0200
> 
>     block: fix locking in bdev_del_partition
>     
>     [ Upstream commit 08fc1ab6d748ab1a690fd483f41e2938984ce353 ]
>     
>     We need to hold the whole device bd_mutex to protect against
>     other thread concurrently deleting out partition before we get
>     to it, and thus causing a use after free.
>     
>     Fixes: cddae808aeb7 ("block: pass a hd_struct to delete_partition")
>     Reported-by: syzbot+6448f3c229bc52b82f69@syzkaller.appspotmail.com
>     Signed-off-by: Christoph Hellwig <hch@lst.de>
>     Signed-off-by: Jens Axboe <axboe@kernel.dk>
>     Signed-off-by: Sasha Levin <sashal@kernel.org>
> 
> ?

Do you think that's the faulty commit? I won't have time to test it

today.

Mark H Weaver wrote 4 years ago

Recipients:

Message-ID:87363h9oab.fsf@netris.org

Leo Famulari <leo@famulari.name> writes:

Toggle quote (24 lines)> On Wed, Sep 16, 2020 at 04:52:45PM +0200, Danny Milosavljevic wrote:
>> commit 692d0626557451c4b557397f20b7394b612d0289
>> Author: Christoph Hellwig <hch@lst.de>
>> Date:   Tue Sep 1 11:59:41 2020 +0200
>> 
>>     block: fix locking in bdev_del_partition
>>     
>>     [ Upstream commit 08fc1ab6d748ab1a690fd483f41e2938984ce353 ]
>>     
>>     We need to hold the whole device bd_mutex to protect against
>>     other thread concurrently deleting out partition before we get
>>     to it, and thus causing a use after free.
>>     
>>     Fixes: cddae808aeb7 ("block: pass a hd_struct to delete_partition")
>>     Reported-by: syzbot+6448f3c229bc52b82f69@syzkaller.appspotmail.com
>>     Signed-off-by: Christoph Hellwig <hch@lst.de>
>>     Signed-off-by: Jens Axboe <axboe@kernel.dk>
>>     Signed-off-by: Sasha Levin <sashal@kernel.org>
>> 
>> ?
>
> Do you think that's the faulty commit? I won't have time to test it
> today.

Looks like a good bet to me. Many thanks to Leo for testing 5.8.9, and

to Danny for identifying the likely culprit.

Next, it probably makes sense to test 5.8.9 with the above commit

reverted. Would someone like to try it? If that works, we can avoid

the bisection, resume kernel updates, and revert this change across all

of our kernels until a better solution is found.

Mark

Leo Famulari wrote 4 years ago

Recipients:(name . Mark H Weaver)(address . mhw@netris.org)

Message-ID:20200917022117.GA1845@jasmine.lan

On Wed, Sep 16, 2020 at 09:36:49PM -0400, Mark H Weaver wrote:

Toggle quote (5 lines)

> Next, it probably makes sense to test 5.8.9 with the above commit

> reverted. Would someone like to try it? If that works, we can avoid

> the bisection, resume kernel updates, and revert this change across all

> of our kernels until a better solution is found.

I'll do this tonight.

Marius Bakke wrote 4 years ago

Recipients:

Message-ID:87r1r06bko.fsf@gnu.org

Leo Famulari <leo@famulari.name> writes:

Toggle quote (8 lines)

> On Wed, Sep 16, 2020 at 09:36:49PM -0400, Mark H Weaver wrote:

>> Next, it probably makes sense to test 5.8.9 with the above commit

>> reverted. Would someone like to try it? If that works, we can avoid

>> the bisection, resume kernel updates, and revert this change across all

>> of our kernels until a better solution is found.

> I'll do this tonight.

A fix for this is available in mainline:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=88ce2a530cc9865a894454b2e40eba5957a60e1a

(via https://gitlab.gnome.org/GNOME/gparted/-/issues/111)

Leo Famulari wrote 4 years ago

Recipients:(name . Mark H Weaver)(address . mhw@netris.org)

Message-ID:20200917140544.GA14066@jasmine.lan

On Wed, Sep 16, 2020 at 09:36:49PM -0400, Mark H Weaver wrote:

Toggle quote (12 lines)

> > On Wed, Sep 16, 2020 at 04:52:45PM +0200, Danny Milosavljevic wrote:

> >> commit 692d0626557451c4b557397f20b7394b612d0289

> >> Author: Christoph Hellwig <hch@lst.de>

> >> Date: Tue Sep 1 11:59:41 2020 +0200

> >>

> >> block: fix locking in bdev_del_partition

> Next, it probably makes sense to test 5.8.9 with the above commit

> reverted. Would someone like to try it? If that works, we can avoid

> the bisection, resume kernel updates, and revert this change across all

> of our kernels until a better solution is found.

Using linux-libre 5.8.9 with that commit reverted, `guix system vm` does

work again.

Leo Famulari wrote 4 years ago

Recipients:(name . Marius Bakke)(address . marius@gnu.org)

Message-ID:20200917142026.GB14066@jasmine.lan

On Thu, Sep 17, 2020 at 10:40:07AM +0200, Marius Bakke wrote:

Toggle quote (16 lines)> Leo Famulari <leo@famulari.name> writes:
> 
> > On Wed, Sep 16, 2020 at 09:36:49PM -0400, Mark H Weaver wrote:
> >> Next, it probably makes sense to test 5.8.9 with the above commit
> >> reverted.  Would someone like to try it?  If that works, we can avoid
> >> the bisection, resume kernel updates, and revert this change across all
> >> of our kernels until a better solution is found.
> >
> > I'll do this tonight.
> 
> A fix for this is available in mainline:
> 
>   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=88ce2a530cc9865a894454b2e40eba5957a60e1a
> 
> (via <https://gitlab.gnome.org/GNOME/gparted/-/issues/111>)

What is the recommended way to report this to the stable maintainers?

Danny Milosavljevic wrote 4 years ago

Recipients:(name . Leo Famulari)(address . leo@famulari.name)

Message-ID:20200917172806.1d03de5c@scratchpost.org

Hi Leo,

On Thu, 17 Sep 2020 10:05:44 -0400

Leo Famulari <leo@famulari.name> wrote:

Toggle quote (16 lines)> On Wed, Sep 16, 2020 at 09:36:49PM -0400, Mark H Weaver wrote:
> > > On Wed, Sep 16, 2020 at 04:52:45PM +0200, Danny Milosavljevic wrote:  
> > >> commit 692d0626557451c4b557397f20b7394b612d0289
> > >> Author: Christoph Hellwig <hch@lst.de>
> > >> Date:   Tue Sep 1 11:59:41 2020 +0200
> > >> 
> > >>     block: fix locking in bdev_del_partition  
> 
> > Next, it probably makes sense to test 5.8.9 with the above commit
> > reverted.  Would someone like to try it?  If that works, we can avoid
> > the bisection, resume kernel updates, and revert this change across all
> > of our kernels until a better solution is found.  
> 
> Using linux-libre 5.8.9 with that commit reverted, `guix system vm` does
> work again.

Thanks for testing it!

What the commit does is lock the drive whose partition table is being

modified. That sounds like a good idea in general. But parted does not

expect the error code ENOMEM and thus misjudges the situation.

is:open	open issues
is:done	closed issues
submitter:<who>	search issue submitter
author:<who>	search by message author
date:yesterday..now	search by issue date
mdate:3m..2d	search by message date