| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| In the Linux kernel, the following vulnerability has been resolved:
memcg: avoid charging the root memcg from obj_cgroup_charge_pages()
obj_cgroup_charge_pages() resolves the objcg to its memcg and calls
try_charge_memcg(), which does not short circuit the root memcg. That
memcg can be the root memcg: obj_cgroup_is_root() reflects the memcg the
objcg was created for and is never updated, while memcg_reparent_objcgs()
does redirect objcg->memcg to the parent on rmdir. An objcg of a dying
child of root therefore passes every obj_cgroup_is_root() filter but
resolves to the root memcg.
Folios keep the objcg they were charged with, so this is easy to reach
through zswap: allocate anon memory in a cgroup, move the task out, remove
the cgroup, then write to the root cgroup's memory.reclaim. The reclaimed
folios are charged through the reparented objcg and end up in
refill_stock() with the root memcg:
WARNING: mm/memcontrol.c:2198 at refill_stock+0x644/0x940
refill_stock+0x644/0x940
try_charge_memcg+0x12d6/0x1570
__obj_cgroup_charge+0x35/0xf0
obj_cgroup_charge+0x1de/0x210
obj_cgroup_charge_zswap+0x83/0x270
zswap_store+0x1620/0x2000
swap_writeout+0x94c/0x14c0
shrink_folio_list+0x3388/0x52b0
[...]
try_to_free_mem_cgroup_pages+0x30d/0x830
user_proactive_reclaim+0x504/0x840
memory_reclaim+0x1f/0x30
Beyond the warning, the charge is asymmetric: obj_cgroup_uncharge_pages()
skips refill_stock() for the root memcg, so the root's page counter grows
and is never uncharged. It is not user visible, since memory.current is
not exposed on the root, but it is a leak.
Use try_charge(), which returns early for the root memcg, restoring the
symmetry with obj_cgroup_uncharge_pages().
The above sequence was scripted into a standalone reproducer (zswap on,
swap on a virtio disk, 512MB of anon memory faulted in inside a child of
the root cgroup, the task then migrated to the root cgroup, the child
removed, followed by "echo 600M swappiness=max > memory.reclaim" on the
root) and run in a CONFIG_DEBUG_VM=y VM. It reproduces the splat on the
first zswap store of a reparented folio, with the same call chain as the
report. With this patch applied the splat is gone while the zswap store
count over the run is unchanged, so the same path is still exercised.
cgroup selftests test_zswap, test_kmem and test_memcontrol show no new
failures. |
| In the Linux kernel, the following vulnerability has been resolved:
selinux: always fill AVC decision in avc_has_perm_noaudit()
avc_has_perm_noaudit() is documented to return a copy of the access
decision in @avd, but its early return for an empty requested permission
set leaves the buffer untouched. All callers pass an uninitialized
stack variable and afterwards feed it to avc_audit(), and the inode hook
even stores it in the per-task decision cache.
Fill in a deny-all, audit-all decision, similar to avd_init(), so every
caller receives a defined value at no cost on the hot path. |
| In the Linux kernel, the following vulnerability has been resolved:
power: sequencing: fix NULL-pointer dereference in pwrseq_unit_new()
If memory allocation fails in pwrseq_unit_setup_deps(), pwrseq_unit_put()
is called to release the partially initialized unit. However, we've
never initialized unit->list and pwrseq_unit_release() will
unconditionally call list_del() on it. Initialize unit->list right after
allocating the unit struct. |
| In the Linux kernel, the following vulnerability has been resolved:
power: sequencing: fix NULL-pointer dereference in pwrseq_device_register()
If dev_set_name() fails in pwrseq_device_register(), we jump to the
err_put_pwrseq label before initializing pwrseq->targets.
pwrseq_release() will try to iterate over targets unconditionally and
subsequently dereference an invalid pointer. Move the call to
dev_set_name() after the list head is initialized. |
| In the Linux kernel, the following vulnerability has been resolved:
watchdog: da9063: fix suspend/resume handling of HW_RUNNING watchdog
da9063_wdt_suspend() and da9063_wdt_resume() only check watchdog_active(),
when the watchdog is left running by the driver sets
WDOG_HW_RUNNING in da9063_wdt_probe() but userspace never opens the
device, so WDOG_ACTIVE remains cleared, the wdt_disable() will not be
executed in da9063_wdt_suspend. In this case, the suspend callback is
a no-op and the watchdog keeps counting during system suspend,
leading to an unexpected system reset.
Check WDOG_HW_RUNNING and wdd，can fix this issue. |
| In the Linux kernel, the following vulnerability has been resolved:
watchdog: rtd119x: Avoid division by zero
clk_get_rate() could return 0. Avoid a division by zero panic. |
| In the Linux kernel, the following vulnerability has been resolved:
watchdog: rzv2h: Avoid division by zero
clk_get_rate() could return 0. Avoid a division by zero panic. |
| In the Linux kernel, the following vulnerability has been resolved:
wifi: rsi: fix heap OOB write on key removal
When a key is removed (data == NULL), rsi_hal_load_key() runs:
memset(&set_key[FRAME_DESC_SZ], 0, frame_len - FRAME_DESC_SZ);
set_key is a struct rsi_set_key *, so the subscript is scaled by
sizeof(struct rsi_set_key) (160 bytes): &set_key[FRAME_DESC_SZ] is
skb->data + 2560, and the memset writes 144 zero bytes starting
2.4KB past the end of the 160-byte skb data buffer, corrupting
unrelated heap objects. The intended byte offset would have been
(u8 *)set_key + FRAME_DESC_SZ.
The write fires on every DISABLE_KEY callback, so plain disconnects,
roams and interface teardowns trigger it on real networks.
The memset is redundant: the whole buffer is zeroed right after
allocation, so the frame sent to the device is byte-identical
without it. Drop the else branch; normal operation is unaffected.
Discovered by Atuin - Automated Vulnerability Discovery Engine. |
| In the Linux kernel, the following vulnerability has been resolved:
wifi: mac80211: avoid WARN in set_bitrate_mask when sdata not in driver
ieee80211_set_bitrate_mask() checks if the interface is running via
ieee80211_sdata_running(), but it does not check if the interface is
still present in the driver.
When sdata is running but IEEE80211_SDATA_IN_DRIVER is not set, the
call reaches drv_set_bitrate_mask() in driver-ops.h which hits
wlan1: Failed check-sdata-in-driver check, flags: 0x0
WARNING: net/mac80211/driver-ops.h:884 at drv_set_bitrate_mask
Syzkaller triggers this via wext SIOCSIWRATE ioctl. The Call Trace shows
wext_ioctl_dispatch() in wext-core.c dispatching the ioctl, calling
ioctl_standard_call() for SIOCSIWRATE, which calls cfg80211_wext_siwrate()
in wext-compat.c. That builds a bitrate mask and calls
rdev_set_bitrate_mask() which ends up in ieee80211_set_bitrate_mask() in
cfg.c. The interface is marked running via SDATA_STATE_RUNNING but
flags is 0, so check_sdata_in_driver() fails.
When the interface is being torn down, or when wext ioctl is issued
during interface bringup before drv_add_interface() sets IN_DRIVER, the
running check passes while IN_DRIVER is clear.
Check IEEE80211_SDATA_IN_DRIVER in ieee80211_set_bitrate_mask() before
calling the driver, returning -ENETDOWN. This avoids the WARN_ONCE in
driver-ops.h and matches other cfg.c operations that bail early when not
in driver.
This change should be safe because wiphy mutex is held in
cfg80211_wext_siwrate() via guard(wiphy), and IN_DRIVER is set/cleared
under RTNL and wiphy paths in drv_add_interface() and
drv_remove_interface() in driver-ops.c, so the check is race-free
against driver add/remove. Returning -ENETDOWN is the same error other
not-running paths use and does not introduce new locking.
[also add second syzbot report] |
| In the Linux kernel, the following vulnerability has been resolved:
smb: client: fix missing iov bounds check in parse_posix_sids()
In parse_posix_sids(), sidsbuf_end is calculated using the server-supplied
out_len without being validated against the actual length of the received
iov (iov_len).
If a server provides an inflated out_len, sidsbuf_end will point past the
end of the iov. This defeats the bounds guards in posix_info_sid_size(),
allowing out-of-bounds reads into adjacent kernel memory.
Fix this by rejecting responses where the calculated sidsbuf_end would
exceed the received iov boundaries or cause pointer wraparound. |
| In the Linux kernel, the following vulnerability has been resolved:
drm/virtio: Fix a NULL vs ERR_PTR() bug in virtio_gpu_user_framebuffer_create()
Smatch complains that returning a NULL here will lead to a NULL pointer
dereference in drm_mode_addfb2(). Return an error pointer instead. |
| In the Linux kernel, the following vulnerability has been resolved:
s390/boot: Avoid IPL parameter append past command line
A command line may occupy all but the terminating byte of
COMMAND_LINE_SIZE. In that case append_ipl_block_parm() passes a zero size
to the IPL parameter conversion helpers and points the destination one
byte past early_command_line. The helpers subtract one from the unsigned
size and write the converted parameter outside the command line buffer.
Convert the IPL parameter in the command line parsing buffer first. A
parameter beginning with '=' can then replace the existing command line
regardless of its length, while other parameters are appended only when
space remains. |
| In the Linux kernel, the following vulnerability has been resolved:
net: icmp: avoid invalid transport header access in icmp_send tracepoint
syzbot reported a WARNING triggered by DEBUG_NET_WARN_ON_ONCE():
WARNING: at skb_transport_header include/linux/skbuff.h:3087 [inline]
WARNING: at udp_hdr include/linux/udp.h:23 [inline]
WARNING: at do_trace_event_raw_event_icmp_send include/trace/events/icmp.h:30 [inline]
WARNING: at trace_event_raw_event_icmp_send+0x48c/0x6ec include/trace/events/icmp.h:11
Call trace:
skb_transport_header include/linux/skbuff.h:3087 [inline]
udp_hdr include/linux/udp.h:23 [inline]
do_trace_event_raw_event_icmp_send include/trace/events/icmp.h:30 [inline]
trace_event_raw_event_icmp_send+0x48c/0x6ec include/trace/events/icmp.h:11
__traceiter_icmp_send include/trace/events/icmp.h:11 [inline]
__do_trace_icmp_send include/trace/events/icmp.h:11 [inline]
trace_icmp_send+0x320/0x49c include/trace/events/icmp.h:11
__icmp_send+0xcfc/0x11d8 net/ipv4/icmp.c:1013
ipv4_send_dest_unreach net/ipv4/route.c:1280 [inline]
ipv4_link_failure+0x57c/0x8dc net/ipv4/route.c:1287
dst_link_failure include/net/dst.h:438 [inline]
vti_tunnel_xmit+0xe40/0x17a4 net/ipv4/ip_vti.c:307
TP_fast_assign() unconditionally calls udp_hdr(skb) before checking
whether the packet is UDP. Furthermore, __icmp_send() can be invoked
from paths (e.g., link failures, ARP errors, forwarding, AF_PACKET)
where skb->transport_header was never initialized (~0U).
Under CONFIG_DEBUG_NET=y, calling skb_transport_header(skb) triggers
DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb)).
Fix this by:
1. Only parsing transport info when iph->protocol == IPPROTO_UDP.
2. Using skb_header_pointer() at skb_network_offset(skb) + (iph->ihl << 2)
to safely fetch the UDP header without assuming transport_header is set. |
| In the Linux kernel, the following vulnerability has been resolved:
net: iptunnel: fix stale transport header during tunnel decapsulation
Syzbot reported a crash in qdisc_pkt_len_segs_init() caused by a stale
transport_header offset after tunnel decapsulation.
BUG: unable to handle page fault for address: ffffed102091a42e
Oops: Oops: 0000 [#1] SMP KASAN NOPTI
CPU: 0 UID: 0 PID: 340 Comm: qdisc_uaf_repro Not tainted 7.2.0-rc4-00061-g248951ddc14d #256 PREEMPT(full)
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
RIP: 0010:__asan_load2
<IRQ>
qdisc_pkt_len_segs_init (net/core/dev.c:4145)
__dev_queue_xmit (net/core/dev.c:4787)
br_dev_queue_push_xmit (net/bridge/br_forward.c:53)
br_handle_frame_finish (net/bridge/br_input.c:229)
br_handle_frame (net/bridge/br_input.c:315)
__netif_receive_skb_core.constprop.0 (net/core/dev.c:6099)
__netif_receive_skb_list_core (net/core/dev.c:6287)
netif_receive_skb_list_internal (net/core/dev.c:6445)
napi_complete_done (net/core/dev.c:6813)
gro_cell_poll (net/core/gro_cells.c:74)
__napi_poll (net/core/dev.c:7735)
net_rx_action (net/core/dev.c:7798 net/core/dev.c:7955)
handle_softirqs (kernel/softirq.c:622)
do_softirq (kernel/softirq.c:523 kernel/softirq.c:510 )
__local_bh_enable_ip (kernel/softirq.c:450)
tun_get_user (drivers/net/tun.c:1986 (discriminator 1))
tun_chr_write_iter (drivers/net/tun.c:2032)
The issue is completely latent until qdisc read transport header in
commit 7fb4c1967011 ("net: pull headers in qdisc_pkt_len_segs_init()").
The crash requires four conditions to line up:
1. The incoming packet is encapsulated and carries GSO metadata. The outer
transport header offset is stored in skb->transport_header while the
packet is still in the outer tunnel context.
2. The tunnel receiver strips the outer headers. skb->data is advanced to
the inner frame, but skb->transport_header is left pointing to the
now-removed outer L4 header, so it becomes a negative offset relative to
the new data.
3. The inner frame is not delivered to the local IP stack. Instead, it
is forwarded at L2 by a bridge or HSR, so ip_rcv_core() never runs and
the transport header is not reset to the inner L4 offset.
4. The forwarding path calls __dev_queue_xmit(), which enters
qdisc_pkt_len_segs_init(). That function computes the GSO header length
from skb_transport_offset(skb). Because the offset is negative, the
unsigned cast overflows and pskb_may_pull(skb, hdr_len +
sizeof(struct tcphdr)) reads past the end of the skb, triggering a
KASAN fault or page fault.
The issue specifically requires GSO packets (shinfo->gso_size != 0), which
are processed/aggregated through gro_cells. Fix this by clearing
transport_header to the ~0U sentinel in gro_cell for all tunnnel driver.
GTP does not support GRO/GSO, drop the evil GSO packets in GTP directly. |
| In the Linux kernel, the following vulnerability has been resolved:
s390/ipl: Fix NULL deref in kdump without re-IPL parm block
Some IPL types, like HMC FTP boot or QEMU direct kernel boot, might
not provide an IPL parameter block. In this case, reipl_type_init()
selects IPL_TYPE_UNKNOWN, and reipl_block_actual remains NULL.
kdump passes the re-IPL parameter block to the dump kernel through
os_info. Before commit 3b9678472bab ("s390/ipl: correct kdump reipl
block checksum calculation"), the os_info entry was added only for
IPL types which initialized reipl_block_actual. That commit moved the
os_info update to machine_crash_shutdown(), making it unconditional. As
a result, set_os_info_reipl_block() dereferences reipl_block_actual for
IPL_TYPE_UNKNOWN. This may happen to work by chance when address zero
contains readable lowcore data and the resulting empty os_info entry is
ignored by the dump kernel.
Skip the os_info update when no re-IPL parameter block is available.
Kdump then collect the dump and reboot without setting re-IPL parameter
block. |
| In the Linux kernel, the following vulnerability has been resolved:
s390/ipl: Fix NULL deref in dump_reipl without re-IPL parm block
Unlike kdump, which passes the re-IPL parameter block through os_info,
the stand-alone dump passes it through the IPL parm block address and
checksum in lowcore.
Some IPL types, like HMC FTP boot or QEMU direct kernel boot, might not
provide an IPL parameter block. In this case reipl_type_init() selects
IPL_TYPE_UNKNOWN and reipl_block_actual remains NULL. Nevertheless,
dump_reipl_run() unconditionally dereferences it when preparing the
lowcore fields. This may happen to work by chance when address zero
contains readable lowcore data. A zero IPL parameter block address is
then stored in lowcore, causing the stand-alone dumper to enter disabled
wait after completing the dump.
Explicitly store a zero IPL parameter block address and checksum when no
re-IPL parameter block is available. This does not change the behavior:
the stand-alone dumper completes the dump and halts, while valid re-IPL
parameter blocks continue to be handled as before. |
| In the Linux kernel, the following vulnerability has been resolved:
locking/lockdep: Invalidate stale class_cache entries for zapped classes
syzbot reported a lockdep splat hitting DEBUG_LOCKS_WARN_ON(1) in
hlock_class() due to an invalid class_idx:
WARNING: kernel/locking/lockdep.c:238 at __lock_acquire+0x382/0x2cf0 kernel/locking/lockdep.c:5203
Workqueue: wg-crypt-wg0 wg_packet_tx_worker
RIP: 0010:hlock_class kernel/locking/lockdep.c:238 [inline]
RIP: 0010:check_wait_context kernel/locking/lockdep.c:4870 [inline]
RIP: 0010:__lock_acquire+0x389/0x2cf0 kernel/locking/lockdep.c:5203
Call Trace:
<IRQ>
lock_acquire+0x106/0x350 kernel/locking/lockdep.c:5886
_raw_spin_lock+0x2e/0x40 kernel/locking/spinlock.c:173
tcp_tsq_handler+0x29/0x200 net/ipv4/tcp_output.c:1291
tcp_tsq_workfn+0x384/0x410 net/ipv4/tcp_output.c:1325
...
When a lock class is zapped (e.g. during module unload or key
unregistration), zap_class() clears the class's bit in
lock_classes_in_use and removes it from the class hash table.
However, existing lockdep_map instances embedded in data structures
may still retain a pointer to the zapped class in their class_cache[]
array.
When __lock_acquire() subsequently runs on such a lock, it finds
lock->class_cache[subclass] != NULL, skipping register_lock_class()
and assigning hlock->class_idx to the index of the zapped class. When
check_wait_context() or hlock_class() inspects the held_lock, it finds
!test_bit(class_idx, lock_classes_in_use) and warns. Furthermore, if
the zapped slot is subsequently re-allocated to an unrelated lock key,
the stale class_cache entry would erroneously match the unrelated
class (ABA issue).
Add lock_class_cache_is_valid() to validate that the cached class is
within lock_classes bounds, still allocated in lock_classes_in_use
(using uninstrumented arch_test_bit() in __always_inline context so it
is safe in noinstr contexts like match_held_lock()), and that
class->key matches the expected subkey (taking lockdep_set_subclass()
overrides into account). Also use READ_ONCE()/WRITE_ONCE() when
accessing class_cache[]. If the entry is invalid or stale, fall back
to register_lock_class() / look_up_lock_class(). |
| In the Linux kernel, the following vulnerability has been resolved:
platform/x86: x86-android-tablets: hold device reference for secondary fwnode teardown
In gpio_secondary_fwnode_init(), acpi_bus_find_device_by_name() returns a
device reference, but the local dev variable is declared with
__free(put_device), dropping the reference at the end of each iteration.
Meanwhile, devm_add_action_or_reset() saves the dev pointer for
gpio_secondary_unset() without incrementing its reference count, which
could lead to a use-after-free during driver teardown if the device is
released in the interim.
Acquire an explicit device reference with get_device() when registering
the devres action, and drop it with put_device() inside
gpio_secondary_unset(). |
| In the Linux kernel, the following vulnerability has been resolved:
powerpc: Do not restore KUAP in arch_exit_to_user_mode_prepare()
KUAP means kernel cannot touch user memory unless it explicitly is
enabled. In the kernel it should stay AMR_KUAP_BLOCKED. While returning
to userspace just before RFI, kernel should restore the user AMR value
back.
Looks like GENERIC_ENTRY might be treating arch_exit_to_user_mode_prepare()
as the last architecture step before returning to userspace.
commit bee25f97ad24 ("powerpc: Enable GENERIC_ENTRY feature")
therefore called kuap_user_restore() from that hook. But on PowerPC that
is too early. After irqentry_exit() / syscall_exit_to_user_mode() we
still run platform specific exit routines.
e.g. code snippets showing both exception handling and system call
handling as the callers of function arch_exit_to_user_mode_prepare()
which does kuap_user_restore(). The below path shows that calling
kuap_user_restore() is too early when called from
arch_exit_to_user_mode_prepare().
Exception handling in exceptions-64s.S
=======================================
bl CFUNC(do_page_fault)
..DEFINE_INTERRUPT_HANDLER_ASYNC(do_page_fault)
arch_interrupt_async_enter_prepare(regs);
state = irqentry_enter(regs);
instrumentation_begin();
irq_enter_rcu();
handler(regs);
nap_adjust_return(regs);
irq_exit_rcu();
instrumentation_end();
arch_interrupt_async_exit_prepare(regs);
irqentry_exit(regs, state); <<< too early
irqentry_exit_to_user_mode()
__exit_to_user_mode_prepare(regs, EXIT_TO_USER_MODE_WORK_IRQ);
arch_exit_to_user_mode_prepare(regs, ti_work); <<< too early
b interrupt_return_srr
.. bl CFUNC(interrupt_exit_user_prepare) <<< already calls kuap_user_restore
prep_irq_for_enabled_exit() retry can run kernel code with IRQs on. So
only when that routine is fully finished is when the user KUAP should be
fully restored which interrupt_exit_user_prepare() already takes care of
before returning.
Similarly for system call handling in interrupt_64.S
======================================================
bl CFUNC(system_call_exception)
.Lsyscall_exit:
addi r4,r1,STACK_INT_FRAME_REGS
li r5,0 /* !scv */
bl CFUNC(syscall_exit_prepare)
.. kuap_assert_locked();
syscall_exit_to_user_mode(regs); <<< too early
syscall_exit_to_user_mode_prepare(regs); <<< too early
kuap_user_restore(regs); <<< already calls
syscall_exit_prepare(), which can enable IRQs, replay a pending
interrupt, and only then rfi. Those functions already restore KUAP
immediately before rfi.
Note that if we restore the user AMR too early like in the current code
as shown from the code snippets above, then we get the following warning
when CONFIG_PPC_KUAP_DEBUG is enabled:
WARNING: arch/powerpc/include/asm/book3s/64/kup.h:293 at interrupt_exit_user_prepare+0x1a0/0x1c0
Hardware name: IBM pSeries (emulated by qemu) POWER10 (architected)
TRAP: 0700
LR: c00000000000d8d4 CTR: c0000000021fe500
MSR: <SF,EE,ME,IR,DR,RI,LE> CR: 44000804 XER: 20040000
interrupt_exit_user_prepare+0x1a0/0x1c0
interrupt_return_srr_user+0x8/0x12c |
| In the Linux kernel, the following vulnerability has been resolved:
powerpc/entry: Fix irq_soft_mask corruption on replayed interrupt exit
When __replay_soft_interrupts() replays a pending interrupt (e.g.
PACA_IRQ_DEC -> timer_interrupt), it calls the handler directly with a
synthetic pt_regs. The DEFINE_INTERRUPT_HANDLER_ASYNC wrapper around
each handler calls arch_interrupt_async_exit_prepare() on the way out,
which calls arch_interrupt_exit_prepare() -> local_irq_disable() ->
arch_local_irq_disable(), which does:
irq_soft_mask_set(IRQS_DISABLED) /* 0x1 */
This unconditionally overwrites irq_soft_mask with IRQS_DISABLED (0x1),
stripping the IRQS_PMI_DISABLED (0x2) bit. The result is that
irq_soft_mask is 0x1 instead of IRQS_ALL_DISABLED (0x3) when the
handler returns to __replay_soft_interrupts().
For a normally-taken interrupt this is harmless: the next interrupt
always enters through arch_interrupt_enter_prepare() which
unconditionally sets irq_soft_mask to IRQS_ALL_DISABLED. But during
replay, next_interrupt() is called directly between replayed handlers
without going back through arch_interrupt_enter_prepare(), so the
stripped bit is never restored. next_interrupt() then fires a WARNING:
WARNING: arch/powerpc/kernel/irq_64.c:75
WARN_ON(irq_soft_mask_return() != IRQS_ALL_DISABLED)
The warning was observed early in boot on a POWER10 pseries guest
during kmem_cache_init_late(), where a spinlock release triggers
interrupt replay that processes a pending timer interrupt.
Debugger state confirming the bug:
Before timer_interrupt(®s):
irq_soft_mask = 0x3 (IRQS_ALL_DISABLED) correct
irq_happened = 0x41 (HARD_DIS|REPLAYING) correct
After timer_interrupt(®s) returns:
irq_soft_mask = 0x1 (IRQS_DISABLED) WRONG - PMI bit stripped
irq_happened = 0x41 unchanged
The fix is to replace local_irq_disable() with hard_irq_disable().
hard_irq_disable() is the right primitive here for two reasons:
1. On PPC64 (hw_irq.h:301) it calls irq_soft_mask_set_return(IRQS_ALL_DISABLED),
setting the soft mask to 0x3 (both IRQS_DISABLED and IRQS_PMI_DISABLED),
which preserves the PMI bit and fixes the WARNING. The additional
work it does (__hard_irq_disable(), PACA_IRQ_HARD_DIS |=) is
redundant but safe since both are already set at this point in the
exit path; the trace_hardirqs_off() inside is guarded by
if (!arch_irqs_disabled_flags(flags)) so it will not double-fire.
2. On PPC32 (hw_irq.h:467) hard_irq_disable() maps to
arch_local_irq_disable() -> __hard_irq_disable(), which clears
MSR[EE] in hardware. This is exactly correct: PPC32 has no soft-mask
PACA mechanism, so the hardware disable is the right way to satisfy
irqentry_exit()'s requirement. This also fixes a build error on PPC32
where irq_soft_mask_set() is only defined under CONFIG_PPC64:
arch/powerpc/include/asm/entry-common.h:273: error: implicit
declaration of function 'irq_soft_mask_set'
Using hard_irq_disable() requires no #ifdef and is consistent with
how the rest of the entry code (e.g. entry-common.h:463) handles the
same PPC32/PPC64 split. |
