[PATCH 0/2] printk_ringbuffer: Fix regression in get_data() and clean up data size checks

[Petr Mladek]

This is outcome of the long discussion about the regression caused
by 67e1b0052f6bb82 ("printk_ringbuffer: don't needlessly wrap data blocks around"),
see https://lore.kernel.org/all/69096836.a70a022...@google.com/

The 1st patch fixes the regression as agreed, see
https://lore.kernel.org/all/87ecqb3...@jogness.linutronix.de/

The 2nd patch adds a helper function to unify the checks whether
a more space is needed. I did my best to address all the concerns
about various proposed variants.

Note that I called the new helper function "need_more_space()" in the end.
It avoids all the problems with "before" vs. "lt" vs "le",
and "_safe" vs. "_sane" vs. "_bounded".

IMHO, the name "need_more_space()" fits very well in all three
locations, surprisingly even in data_realloc(). But it is possible
that you disagree. Let me know if you hate it ;-)


The patchset applies on top of printk/linux.git, branch for-6.19.
It should apply on top of linux-next as well.

Petr Mladek (2):
printk_ringbuffer: Fix check of valid data size when blk_lpos
overflows
printk_ringbuffer: Create a helper function to decide whether a more
space is needed

kernel/printk/printk_ringbuffer.c | 40 +++++++++++++++++++++++++------
1 file changed, 33 insertions(+), 7 deletions(-)

--
2.51.1

[Petr Mladek]

The commit 67e1b0052f6bb8 ("printk_ringbuffer: don't needlessly wrap
data blocks around") allows to use the last 4 bytes of the ring buffer.

But the check for the @data_size was not properly updated in get_data().
It fails when "blk_lpos->next" overflows to "0". In this case:

+ is_blk_wrapped(data_ring, blk_lpos->begin, blk_lpos->next)
returns "false" because it checks "blk_lpos->next - 1".

+ "blk_lpos->begin < blk_lpos->next" fails because "blk_lpos->next"
is already 0.

+ is_blk_wrapped(data_ring, blk_lpos->begin + DATA_SIZE(data_ring),
blk_lpos->next) returns "false" because "begin_lpos" is from
the next wrap but "next_lpos - 1" is from the previous one.

As a result, get_data() triggers the WARN_ON_ONCE() for "Illegal
block description", for example:

[ 216.317316][ T7652] loop0: detected capacity change from 0 to 16
** 1 printk messages dropped **
[ 216.327750][ T7652] ------------[ cut here ]------------
[ 216.327789][ T7652] WARNING: kernel/printk/printk_ringbuffer.c:1278 at get_data+0x48a/0x840, CPU#1: syz.0.585/7652
[ 216.327848][ T7652] Modules linked in:
[ 216.327907][ T7652] CPU: 1 UID: 0 PID: 7652 Comm: syz.0.585 Not tainted syzkaller #0 PREEMPT(full)
[ 216.327933][ T7652] Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/02/2025
[ 216.327953][ T7652] RIP: 0010:get_data+0x48a/0x840
[ 216.327986][ T7652] Code: 83 c4 f8 48 b8 00 00 00 00 00 fc ff df 41 0f b6 04 07 84 c0 0f 85 ee 01 00 00 44 89 65 00 49 83 c5 08 eb 13 e8 a7 19 1f 00 90 <0f> 0b 90 eb 05 e8 9c 19 1f 00 45 31 ed 4c 89 e8 48 83 c4 28 5b 41
[ 216.328007][ T7652] RSP: 0018:ffffc900035170e0 EFLAGS: 00010293
[ 216.328029][ T7652] RAX: ffffffff81a1eee9 RBX: 00003fffffffffff RCX: ffff888033255b80
[ 216.328048][ T7652] RDX: 0000000000000000 RSI: 00003fffffffffff RDI: 0000000000000000
[ 216.328063][ T7652] RBP: 0000000000000012 R08: 0000000000000e55 R09: 000000325e213cc7
[ 216.328079][ T7652] R10: 000000325e213cc7 R11: 00001de4c2000037 R12: 0000000000000012
[ 216.328095][ T7652] R13: 0000000000000000 R14: ffffc90003517228 R15: 1ffffffff1bca646
[ 216.328111][ T7652] FS: 00007f44eb8da6c0(0000) GS:ffff888125fda000(0000) knlGS:0000000000000000
[ 216.328131][ T7652] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 216.328147][ T7652] CR2: 00007f44ea9722e0 CR3: 0000000066344000 CR4: 00000000003526f0
[ 216.328168][ T7652] Call Trace:
[ 216.328178][ T7652] <TASK>
[ 216.328199][ T7652] _prb_read_valid+0x672/0xa90
[ 216.328328][ T7652] ? desc_read+0x1b8/0x3f0
[ 216.328381][ T7652] ? __pfx__prb_read_valid+0x10/0x10
[ 216.328422][ T7652] ? panic_on_this_cpu+0x32/0x40
[ 216.328450][ T7652] prb_read_valid+0x3c/0x60
[ 216.328482][ T7652] printk_get_next_message+0x15c/0x7b0
[ 216.328526][ T7652] ? __pfx_printk_get_next_message+0x10/0x10
[ 216.328561][ T7652] ? __lock_acquire+0xab9/0xd20
[ 216.328595][ T7652] ? console_flush_all+0x131/0xb10
[ 216.328621][ T7652] ? console_flush_all+0x478/0xb10
[ 216.328648][ T7652] console_flush_all+0x4cc/0xb10
[ 216.328673][ T7652] ? console_flush_all+0x131/0xb10
[ 216.328704][ T7652] ? __pfx_console_flush_all+0x10/0x10
[ 216.328748][ T7652] ? is_printk_cpu_sync_owner+0x32/0x40
[ 216.328781][ T7652] console_unlock+0xbb/0x190
[ 216.328815][ T7652] ? __pfx___down_trylock_console_sem+0x10/0x10
[ 216.328853][ T7652] ? __pfx_console_unlock+0x10/0x10
[ 216.328899][ T7652] vprintk_emit+0x4c5/0x590
[ 216.328935][ T7652] ? __pfx_vprintk_emit+0x10/0x10
[ 216.328993][ T7652] _printk+0xcf/0x120
[ 216.329028][ T7652] ? __pfx__printk+0x10/0x10
[ 216.329051][ T7652] ? kernfs_get+0x5a/0x90
[ 216.329090][ T7652] _erofs_printk+0x349/0x410
[ 216.329130][ T7652] ? __pfx__erofs_printk+0x10/0x10
[ 216.329161][ T7652] ? __raw_spin_lock_init+0x45/0x100
[ 216.329186][ T7652] ? __init_swait_queue_head+0xa9/0x150
[ 216.329231][ T7652] erofs_fc_fill_super+0x1591/0x1b20
[ 216.329285][ T7652] ? __pfx_erofs_fc_fill_super+0x10/0x10
[ 216.329324][ T7652] ? sb_set_blocksize+0x104/0x180
[ 216.329356][ T7652] ? setup_bdev_super+0x4c1/0x5b0
[ 216.329385][ T7652] get_tree_bdev_flags+0x40e/0x4d0
[ 216.329410][ T7652] ? __pfx_erofs_fc_fill_super+0x10/0x10
[ 216.329444][ T7652] ? __pfx_get_tree_bdev_flags+0x10/0x10
[ 216.329483][ T7652] vfs_get_tree+0x92/0x2b0
[ 216.329512][ T7652] do_new_mount+0x302/0xa10
[ 216.329537][ T7652] ? apparmor_capable+0x137/0x1b0
[ 216.329576][ T7652] ? __pfx_do_new_mount+0x10/0x10
[ 216.329605][ T7652] ? ns_capable+0x8a/0xf0
[ 216.329637][ T7652] ? kmem_cache_free+0x19b/0x690
[ 216.329682][ T7652] __se_sys_mount+0x313/0x410
[ 216.329717][ T7652] ? __pfx___se_sys_mount+0x10/0x10
[ 216.329836][ T7652] ? do_syscall_64+0xbe/0xfa0
[ 216.329869][ T7652] ? __x64_sys_mount+0x20/0xc0
[ 216.329901][ T7652] do_syscall_64+0xfa/0xfa0
[ 216.329932][ T7652] ? lockdep_hardirqs_on+0x9c/0x150
[ 216.329964][ T7652] ? entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 216.329988][ T7652] ? clear_bhb_loop+0x60/0xb0
[ 216.330017][ T7652] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 216.330040][ T7652] RIP: 0033:0x7f44ea99076a
[ 216.330080][ T7652] Code: d8 64 89 02 48 c7 c0 ff ff ff ff eb a6 e8 de 1a 00 00 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 49 89 ca b8 a5 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
[ 216.330100][ T7652] RSP: 002b:00007f44eb8d9e68 EFLAGS: 00000246 ORIG_RAX: 00000000000000a5
[ 216.330128][ T7652] RAX: ffffffffffffffda RBX: 00007f44eb8d9ef0 RCX: 00007f44ea99076a
[ 216.330146][ T7652] RDX: 0000200000000180 RSI: 00002000000001c0 RDI: 00007f44eb8d9eb0
[ 216.330164][ T7652] RBP: 0000200000000180 R08: 00007f44eb8d9ef0 R09: 0000000000000000
[ 216.330181][ T7652] R10: 0000000000000000 R11: 0000000000000246 R12: 00002000000001c0
[ 216.330196][ T7652] R13: 00007f44eb8d9eb0 R14: 00000000000001a1 R15: 0000200000000080
[ 216.330233][ T7652] </TASK>

Solve the problem by moving and fixing the sanity check. The problematic
if-else-if-else code will just distinguish three basic scenarios:
"regular" vs. "wrapped" vs. "too many times wrapped" block.

The new sanity check is more precise. A valid "data_size" must be
lower than half of the data buffer size. Also it must not be zero at
this stage. It allows to catch problematic "data_size" even for wrapped
blocks.

Closes: https://lore.kernel.org/all/69096836.a70a022...@google.com/
Closes: https://lore.kernel.org/all/69078fb6.050a022...@google.com/
Fixes: 67e1b0052f6bb82 ("printk_ringbuffer: don't needlessly wrap data blocks around")
Signed-off-by: Petr Mladek <pml...@suse.com>
---
kernel/printk/printk_ringbuffer.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/kernel/printk/printk_ringbuffer.c b/kernel/printk/printk_ringbuffer.c
index 839f504db6d3..3e6fd8d6fa9f 100644
--- a/kernel/printk/printk_ringbuffer.c
+++ b/kernel/printk/printk_ringbuffer.c
@@ -1260,9 +1260,8 @@ static const char *get_data(struct prb_data_ring *data_ring,
return NULL;
}

- /* Regular data block: @begin less than @next and in same wrap. */
- if (!is_blk_wrapped(data_ring, blk_lpos->begin, blk_lpos->next) &&
- blk_lpos->begin < blk_lpos->next) {
+ /* Regular data block: @begin and @next in the same wrap. */
+ if (!is_blk_wrapped(data_ring, blk_lpos->begin, blk_lpos->next)) {
db = to_block(data_ring, blk_lpos->begin);
*data_size = blk_lpos->next - blk_lpos->begin;

@@ -1279,6 +1278,10 @@ static const char *get_data(struct prb_data_ring *data_ring,
return NULL;
}

+ /* Sanity check. Data-less blocks were handled earlier. */
+ if (WARN_ON_ONCE(!data_check_size(data_ring, *data_size) || !*data_size))
+ return NULL;
+
/* A valid data block will always be aligned to the ID size. */
if (WARN_ON_ONCE(blk_lpos->begin != ALIGN(blk_lpos->begin, sizeof(db->id))) ||
WARN_ON_ONCE(blk_lpos->next != ALIGN(blk_lpos->next, sizeof(db->id)))) {
--
2.51.1

[Petr Mladek]

The decision whether some more space is needed is tricky in the printk
ring buffer code:

1. The given lpos values might overflow. A subtraction must be used
instead of a simple "lower than" check.

2. Another CPU might reuse the space in the mean time. It can be
detected when the subtraction is bigger than DATA_SIZE(data_ring).

3. There is exactly enough space when the result of the subtraction
is zero. But more space is needed when the result is exactly
DATA_SIZE(data_ring).

Add a helper function to make sure that the check is done correctly
in all situations. Also it helps to make the code consistent and
better documented.

Suggested-by: John Ogness <john....@linutronix.de>
Link: https://lore.kernel.org/r/87tsz7i...@jogness.linutronix.de

Signed-off-by: Petr Mladek <pml...@suse.com>
---

kernel/printk/printk_ringbuffer.c | 31 +++++++++++++++++++++++++++----
1 file changed, 27 insertions(+), 4 deletions(-)

diff --git a/kernel/printk/printk_ringbuffer.c b/kernel/printk/printk_ringbuffer.c
index 3e6fd8d6fa9f..ede3039dd041 100644
--- a/kernel/printk/printk_ringbuffer.c
+++ b/kernel/printk/printk_ringbuffer.c
@@ -411,6 +411,23 @@ static bool data_check_size(struct prb_data_ring *data_ring, unsigned int size)
return to_blk_size(size) <= DATA_SIZE(data_ring) / 2;
}

+/*
+ * Compare the current and requested logical position and decide
+ * whether more space needed.
+ *
+ * Return false when @lpos_current is already at or beyond @lpos_target.
+ *
+ * Also return false when the difference between the positions is bigger
+ * than the size of the data buffer. It might happen only when the caller
+ * raced with another CPU(s) which already made and used the space.
+ */
+static bool need_more_space(struct prb_data_ring *data_ring,
+ unsigned long lpos_current,
+ unsigned long lpos_target)
+{
+ return lpos_target - lpos_current - 1 < DATA_SIZE(data_ring);
+}
+
/* Query the state of a descriptor. */
static enum desc_state get_desc_state(unsigned long id,
unsigned long state_val)
@@ -577,7 +594,7 @@ static bool data_make_reusable(struct printk_ringbuffer *rb,
unsigned long id;

/* Loop until @lpos_begin has advanced to or beyond @lpos_end. */
- while ((lpos_end - lpos_begin) - 1 < DATA_SIZE(data_ring)) {
+ while (need_more_space(data_ring, lpos_begin, lpos_end)) {
blk = to_block(data_ring, lpos_begin);

/*
@@ -668,7 +685,7 @@ static bool data_push_tail(struct printk_ringbuffer *rb, unsigned long lpos)
* sees the new tail lpos, any descriptor states that transitioned to
* the reusable state must already be visible.
*/
- while ((lpos - tail_lpos) - 1 < DATA_SIZE(data_ring)) {
+ while (need_more_space(data_ring, tail_lpos, lpos)) {
/*
* Make all descriptors reusable that are associated with
* data blocks before @lpos.
@@ -1148,8 +1165,14 @@ static char *data_realloc(struct printk_ringbuffer *rb, unsigned int size,

next_lpos = get_next_lpos(data_ring, blk_lpos->begin, size);

- /* If the data block does not increase, there is nothing to do. */
- if (head_lpos - next_lpos < DATA_SIZE(data_ring)) {
+ /*
+ * Use the current data block when the size does not increase.
+ *
+ * Note that need_more_space() could never return false here because
+ * the difference between the positions was bigger than the data
+ * buffer size. The data block is reopened and can't get reused.
+ */
+ if (!need_more_space(data_ring, head_lpos, next_lpos)) {
if (wrapped)
blk = to_block(data_ring, 0);
else
--
2.51.1