clk: qcom: ipq6018: drop the CLK_SET_RATE_PARENT flag from PLL clocks

commit 99cd4935cb972d0aafb16838bb2aeadbcaf196ce upstream.

GPLL, NSS crypto PLL clock rates are fixed and shouldn't be scaled based
on the request from dependent clocks. Doing so will result in the
unexpected behaviour. So drop the CLK_SET_RATE_PARENT flag from the PLL
clocks.

Cc: stable@vger.kernel.org
Fixes: d9db07f088af ("clk: qcom: Add ipq6018 Global Clock Controller support")
Signed-off-by: Kathiravan Thirumoorthy <quic_kathirav@quicinc.com>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Link: https://lore.kernel.org/r/20230913-gpll_cleanup-v2-2-c8ceb1a37680@quicinc.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: qcom: ipq8074: drop the CLK_SET_RATE_PARENT flag from PLL clocks

commit e641a070137dd959932c7c222e000d9d941167a2 upstream.

GPLL, NSS crypto PLL clock rates are fixed and shouldn't be scaled based
on the request from dependent clocks. Doing so will result in the
unexpected behaviour. So drop the CLK_SET_RATE_PARENT flag from the PLL
clocks.

Cc: stable@vger.kernel.org
Fixes: b8e7e519625f ("clk: qcom: ipq8074: add remaining PLL’s")
Signed-off-by: Kathiravan Thirumoorthy <quic_kathirav@quicinc.com>
Link: https://lore.kernel.org/r/20230913-gpll_cleanup-v2-1-c8ceb1a37680@quicinc.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

integrity: powerpc: Do not select CA_MACHINE_KEYRING

commit 3edc22655647378dea01900f7b04e017ff96bda9 upstream.

No other platform needs CA_MACHINE_KEYRING, either.

This is policy that should be decided by the administrator, not Kconfig
dependencies.

Cc: stable@vger.kernel.org # v6.6+
Fixes: d7d91c4743c4 ("integrity: PowerVM machine keyring enablement")
Signed-off-by: Michal Suchanek <msuchanek@suse.de>
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: visconti: Fix undefined behavior bug in struct visconti_pll_provider

commit 5ad1e217a2b23aa046b241183bd9452d259d70d0 upstream.

`struct clk_hw_onecell_data` is a flexible structure, which means that
it contains flexible-array member at the bottom, in this case array
`hws`:

include/linux/clk-provider.h:
1380 struct clk_hw_onecell_data {
1381         unsigned int num;
1382         struct clk_hw *hws[] __counted_by(num);
1383 };

This could potentially lead to an overwrite of the objects following
`clk_data` in `struct visconti_pll_provider`, in this case
`struct device_node *node;`, at run-time:

drivers/clk/visconti/pll.h:
 16 struct visconti_pll_provider {
 17         void __iomem *reg_base;
 18         struct clk_hw_onecell_data clk_data;
 19         struct device_node *node;
 20 };

Notice that a total of 56 bytes are allocated for flexible-array `hws`
at line 328. See below:

include/dt-bindings/clock/toshiba,tmpv770x.h:
 14 #define TMPV770X_NR_PLL 7

drivers/clk/visconti/pll-tmpv770x.c:
 69 ctx = visconti_init_pll(np, reg_base, TMPV770X_NR_PLL);

drivers/clk/visconti/pll.c:
321 struct visconti_pll_provider * __init visconti_init_pll(struct device_node *np,
322                                                         void __iomem *base,
323                                                         unsigned long nr_plls)
324 {
325         struct visconti_pll_provider *ctx;
...
328         ctx = kzalloc(struct_size(ctx, clk_data.hws, nr_plls), GFP_KERNEL);

`struct_size(ctx, clk_data.hws, nr_plls)` above translates to
sizeof(struct visconti_pll_provider) + sizeof(struct clk_hw *) * 7 ==
24 + 8 * 7 == 24 + 56
  ^^^^
   |
allocated bytes for flex array `hws`

$ pahole -C visconti_pll_provider drivers/clk/visconti/pll.o
struct visconti_pll_provider {
void *                     reg_base;             /*     0     8 */
struct clk_hw_onecell_data clk_data;             /*     8     8 */
struct device_node *       node;                 /*    16     8 */

/* size: 24, cachelines: 1, members: 3 */
/* last cacheline: 24 bytes */
};

And then, after the allocation, some data is written into all members
of `struct visconti_pll_provider`:

332         for (i = 0; i < nr_plls; ++i)
333                 ctx->clk_data.hws[i] = ERR_PTR(-ENOENT);
334
335         ctx->node = np;
336         ctx->reg_base = base;
337         ctx->clk_data.num = nr_plls;

Fix all these by placing the declaration of object `clk_data` at the
end of `struct visconti_pll_provider`. Also, add a comment to make it
clear that this object must always be last in the structure, and
prevent this bug from being introduced again in the future.

-Wflex-array-member-not-at-end is coming in GCC-14, and we are getting
ready to enable it globally.

Fixes: b4cbe606dc36 ("clk: visconti: Add support common clock driver and reset driver")
Cc: stable@vger.kernel.org
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Nobuhiro Iwamatsu <nobuhiro1.iwamatsu@toshiba.co.jp>
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/57a831d94ee2b3889b11525d4ad500356f89576f.1697492890.git.gustavoars@kernel.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: socfpga: Fix undefined behavior bug in struct stratix10_clock_data

commit d761bb01c85b22d5b44abe283eb89019693f6595 upstream.

`struct clk_hw_onecell_data` is a flexible structure, which means that
it contains flexible-array member at the bottom, in this case array
`hws`:

include/linux/clk-provider.h:
1380 struct clk_hw_onecell_data {
1381         unsigned int num;
1382         struct clk_hw *hws[] __counted_by(num);
1383 };

This could potentially lead to an overwrite of the objects following
`clk_data` in `struct stratix10_clock_data`, in this case
`void __iomem *base;` at run-time:

drivers/clk/socfpga/stratix10-clk.h:
  9 struct stratix10_clock_data {
 10         struct clk_hw_onecell_data      clk_data;
 11         void __iomem            *base;
 12 };

There are currently three different places where memory is allocated for
`struct stratix10_clock_data`, including the flex-array `hws` in
`struct clk_hw_onecell_data`:

drivers/clk/socfpga/clk-agilex.c:
469         clk_data = devm_kzalloc(dev, struct_size(clk_data, clk_data.hws,
470                                 num_clks), GFP_KERNEL);

drivers/clk/socfpga/clk-agilex.c:
509         clk_data = devm_kzalloc(dev, struct_size(clk_data, clk_data.hws,
510                                 num_clks), GFP_KERNEL);

drivers/clk/socfpga/clk-s10.c:
400         clk_data = devm_kzalloc(dev, struct_size(clk_data, clk_data.hws,
401                                                  num_clks), GFP_KERNEL);

I'll use just one of them to describe the issue. See below.

Notice that a total of 440 bytes are allocated for flexible-array member
`hws` at line 469:

include/dt-bindings/clock/agilex-clock.h:
 70 #define AGILEX_NUM_CLKS 55

drivers/clk/socfpga/clk-agilex.c:
459         struct stratix10_clock_data *clk_data;
460         void __iomem *base;
...
466
467         num_clks = AGILEX_NUM_CLKS;
468
469         clk_data = devm_kzalloc(dev, struct_size(clk_data, clk_data.hws,
470                                 num_clks), GFP_KERNEL);

`struct_size(clk_data, clk_data.hws, num_clks)` above translates to
sizeof(struct stratix10_clock_data) + sizeof(struct clk_hw *) * 55 ==
16 + 8 * 55 == 16 + 440
    ^^^
     |
allocated bytes for flex-array `hws`

474         for (i = 0; i < num_clks; i++)
475                 clk_data->clk_data.hws[i] = ERR_PTR(-ENOENT);
476
477         clk_data->base = base;

and then some data is written into both `hws` and `base` objects.

Fix this by placing the declaration of object `clk_data` at the end of
`struct stratix10_clock_data`. Also, add a comment to make it clear
that this object must always be last in the structure.

-Wflex-array-member-not-at-end is coming in GCC-14, and we are getting
ready to enable it globally.

Fixes: ba7e258425ac ("clk: socfpga: Convert to s10/agilex/n5x to use clk_hw")
Cc: stable@vger.kernel.org
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/1da736106d8e0806aeafa6e471a13ced490eae22.1698117815.git.gustavoars@kernel.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powercap: intel_rapl: Downgrade BIOS locked limits pr_warn() to pr_debug()

commit a60ec4485f1c72dfece365cf95e6de82bdd74300 upstream.

Before the refactoring the pr_warn() only triggered when
someone explicitly tried to write to a BIOS locked limit.
After the refactoring the warning is also triggering during
system resume. The user can't do anything about this so
printing scary warnings doesn't make sense

Keep the printk but make it pr_debug() instead of pr_warn()
to make it clear it's not a serious issue.

Fixes: 9050a9cd5e4c ("powercap: intel_rapl: Cleanup Power Limits support")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: 6.5+ <stable@vger.kernel.org> # 6.5+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cpufreq: stats: Fix buffer overflow detection in trans_stats()

commit ea167a7fc2426f7685c3735e104921c1a20a6d3f upstream.

Commit 3c0897c180c6 ("cpufreq: Use scnprintf() for avoiding potential
buffer overflow") switched from snprintf to the more secure scnprintf
but never updated the exit condition for PAGE_SIZE.

As the commit say and as scnprintf document, what scnprintf returns what
is actually written not counting the '\0' end char. This results in the
case of len exceeding the size, len set to PAGE_SIZE - 1, as it can be
written at max PAGE_SIZE - 1 (as '\0' is not counted)

Because of len is never set to PAGE_SIZE, the function never break early,
never prints the warning and never return -EFBIG.

Fix this by changing the condition to PAGE_SIZE - 1 to correctly trigger
the error.

Cc: 5.10+ <stable@vger.kernel.org> # 5.10+
Fixes: 3c0897c180c6 ("cpufreq: Use scnprintf() for avoiding potential buffer overflow")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc/power: Add power soft-off when running on qemu

commit d0c219472980d15f5cbc5c8aec736848bda3f235 upstream.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # v6.0+
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc/pdc: Add width field to struct pdc_model

commit 6240553b52c475d9fc9674de0521b77e692f3764 upstream.

PDC2.0 specifies the additional PSW-bit field.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc/agp: Use 64-bit LE values in SBA IOMMU PDIR table

commit 86bb854d134f4429feb35d2e05f55c6e036770d2 upstream.

The PDIR table of the System Bus Adapter (SBA) I/O MMU uses 64-bit
little-endian pointers.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # v6.4+
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

pmdomain: imx: Make imx pgc power domain also set the fwnode

commit 374de39d38f97b0e58cfee88da590b2d056ccf7f upstream.

Currently, The imx pgc power domain doesn't set the fwnode
pointer, which results in supply regulator device can't get
consumer imx pgc power domain device from fwnode when creating
a link.

This causes the driver core to instead try to create a link
between the parent gpc device of imx pgc power domain device and
supply regulator device. However, at this point, the gpc device
has already been bound, and the link creation will fail. So adding
the fwnode pointer to the imx pgc power domain device will fix
this issue.

Signed-off-by: Pengfei Li <pengfei.li_1@nxp.com>
Tested-by: Emil Kronborg <emil.kronborg@protonmail.com>
Fixes: 3fb16866b51d ("driver core: fw_devlink: Make cycle detection more robust")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231020185949.537083-1-pengfei.li_1@nxp.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: module: Fix PLT counting when CONFIG_RANDOMIZE_BASE=n

commit d35686444fc80950c731e33a2f6ad4a55822be9b upstream.

The counting of module PLTs has been broken when CONFIG_RANDOMIZE_BASE=n
since commit:

  3e35d303ab7d22c4 ("arm64: module: rework module VA range selection")

Prior to that commit, when CONFIG_RANDOMIZE_BASE=n, the kernel image and
all modules were placed within a 128M region, and no PLTs were necessary
for B or BL. Hence count_plts() and partition_branch_plt_relas() skipped
handling B and BL when CONFIG_RANDOMIZE_BASE=n.

After that commit, modules can be placed anywhere within a 2G window
regardless of CONFIG_RANDOMIZE_BASE, and hence PLTs may be necessary for
B and BL even when CONFIG_RANDOMIZE_BASE=n. Unfortunately that commit
failed to update count_plts() and partition_branch_plt_relas()
accordingly.

Due to this, module_emit_plt_entry() may fail if an insufficient number
of PLT entries have been reserved, resulting in modules failing to load
with -ENOEXEC.

Fix this by counting PLTs regardless of CONFIG_RANDOMIZE_BASE in
count_plts() and partition_branch_plt_relas().

Fixes: 3e35d303ab7d ("arm64: module: rework module VA range selection")
Signed-off-by: Maria Yu <quic_aiquny@quicinc.com>
Cc: <stable@vger.kernel.org> # 6.5.x
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Fixes: 3e35d303ab7d ("arm64: module: rework module VA range selection")
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Link: https://lore.kernel.org/r/20231024010954.6768-1-quic_aiquny@quicinc.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64: Restrict CPU_BIG_ENDIAN to GNU as or LLVM IAS 15.x or newer

commit 146a15b873353f8ac28dc281c139ff611a3c4848 upstream.

Prior to LLVM 15.0.0, LLVM's integrated assembler would incorrectly
byte-swap NOP when compiling for big-endian, and the resulting series of
bytes happened to match the encoding of FNMADD S21, S30, S0, S0.

This went unnoticed until commit:

  34f66c4c4d5518c1 ("arm64: Use a positive cpucap for FP/SIMD")

Prior to that commit, the kernel would always enable the use of FPSIMD
early in boot when __cpu_setup() initialized CPACR_EL1, and so usage of
FNMADD within the kernel was not detected, but could result in the
corruption of user or kernel FPSIMD state.

After that commit, the instructions happen to trap during boot prior to
FPSIMD being detected and enabled, e.g.

| Unhandled 64-bit el1h sync exception on CPU0, ESR 0x000000001fe00000 -- ASIMD
| CPU: 0 PID: 0 Comm: swapper Not tainted 6.6.0-rc3-00013-g34f66c4c4d55 #1
| Hardware name: linux,dummy-virt (DT)
| pstate: 400000c9 (nZcv daIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
| pc : __pi_strcmp+0x1c/0x150
| lr : populate_properties+0xe4/0x254
| sp : ffffd014173d3ad0
| x29: ffffd014173d3af0 x28: fffffbfffddffcb8 x27: 0000000000000000
| x26: 0000000000000058 x25: fffffbfffddfe054 x24: 0000000000000008
| x23: fffffbfffddfe000 x22: fffffbfffddfe000 x21: fffffbfffddfe044
| x20: ffffd014173d3b70 x19: 0000000000000001 x18: 0000000000000005
| x17: 0000000000000010 x16: 0000000000000000 x15: 00000000413e7000
| x14: 0000000000000000 x13: 0000000000001bcc x12: 0000000000000000
| x11: 00000000d00dfeed x10: ffffd414193f2cd0 x9 : 0000000000000000
| x8 : 0101010101010101 x7 : ffffffffffffffc0 x6 : 0000000000000000
| x5 : 0000000000000000 x4 : 0101010101010101 x3 : 000000000000002a
| x2 : 0000000000000001 x1 : ffffd014171f2988 x0 : fffffbfffddffcb8
| Kernel panic - not syncing: Unhandled exception
| CPU: 0 PID: 0 Comm: swapper Not tainted 6.6.0-rc3-00013-g34f66c4c4d55 #1
| Hardware name: linux,dummy-virt (DT)
| Call trace:
|  dump_backtrace+0xec/0x108
|  show_stack+0x18/0x2c
|  dump_stack_lvl+0x50/0x68
|  dump_stack+0x18/0x24
|  panic+0x13c/0x340
|  el1t_64_irq_handler+0x0/0x1c
|  el1_abort+0x0/0x5c
|  el1h_64_sync+0x64/0x68
|  __pi_strcmp+0x1c/0x150
|  unflatten_dt_nodes+0x1e8/0x2d8
|  __unflatten_device_tree+0x5c/0x15c
|  unflatten_device_tree+0x38/0x50
|  setup_arch+0x164/0x1e0
|  start_kernel+0x64/0x38c
|  __primary_switched+0xbc/0xc4

Restrict CONFIG_CPU_BIG_ENDIAN to a known good assembler, which is
either GNU as or LLVM's IAS 15.0.0 and newer, which contains the linked
commit.

Closes: https://github.com/ClangBuiltLinux/linux/issues/1948
Link: https://github.com/llvm/llvm-project/commit/1379b150991f70a5782e9a143c2ba5308da1161c
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Cc: stable@vger.kernel.org
Acked-by: Mark Rutland <mark.rutland@arm.com>
Link: https://lore.kernel.org/r/20231025-disable-arm64-be-ias-b4-llvm-15-v1-1-b25263ed8b23@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

pmdomain: amlogic: Fix mask for the second NNA mem PD domain

commit b131329b9bfbd1b4c0c5e088cb0c6ec03a12930f upstream.

Without this change, the NPU hangs when the 8th NN core is used.

It matches what the out-of-tree driver does.

Signed-off-by: Tomeu Vizoso <tomeu@tomeuvizoso.net>
Fixes: 9a217b7e8953 ("soc: amlogic: meson-pwrc: Add NNA power domain for A311D")
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231016080205.41982-2-tomeu@tomeuvizoso.net
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

PCI: keystone: Don't discard .probe() callback

commit 7994db905c0fd692cf04c527585f08a91b560144 upstream.

The __init annotation makes the ks_pcie_probe() function disappear after
booting completes. However a device can also be bound later. In that case,
we try to call ks_pcie_probe(), but the backing memory is likely already
overwritten.

The right thing to do is do always have the probe callback available.  Note
that the (wrong) __refdata annotation prevented this issue to be noticed by
modpost.

Fixes: 0c4ffcfe1fbc ("PCI: keystone: Add TI Keystone PCIe driver")
Link: https://lore.kernel.org/r/20231001170254.2506508-5-u.kleine-koenig@pengutronix.de
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

PCI: keystone: Don't discard .remove() callback

commit 200bddbb3f5202bbce96444fdc416305de14f547 upstream.

With CONFIG_PCIE_KEYSTONE=y and ks_pcie_remove() marked with __exit, the
function is discarded from the driver. In this case a bound device can
still get unbound, e.g via sysfs. Then no cleanup code is run resulting in
resource leaks or worse.

The right thing to do is do always have the remove callback available.
Note that this driver cannot be compiled as a module, so ks_pcie_remove()
was always discarded before this change and modpost couldn't warn about
this issue. Furthermore the __ref annotation also prevents a warning.

Fixes: 0c4ffcfe1fbc ("PCI: keystone: Add TI Keystone PCIe driver")
Link: https://lore.kernel.org/r/20231001170254.2506508-4-u.kleine-koenig@pengutronix.de
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KEYS: trusted: Rollback init_trusted() consistently

commit 31de287345f41bbfaec36a5c8cbdba035cf76442 upstream.

Do bind neither static calls nor trusted_key_exit() before a successful
init, in order to maintain a consistent state. In addition, depart the
init_trusted() in the case of a real error (i.e. getting back something
else than -ENODEV).

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Closes: https://lore.kernel.org/linux-integrity/CAHk-=whOPoLaWM8S8GgoOPT7a2+nMH5h3TLKtn=R_3w4R1_Uvg@mail.gmail.com/
Cc: stable@vger.kernel.org # v5.13+
Fixes: 5d0682be3189 ("KEYS: trusted: Add generic trusted keys framework")
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KEYS: trusted: tee: Refactor register SHM usage

commit c745cd1718b7825d69315fe7127e2e289e617598 upstream.

The OP-TEE driver using the old SMC based ABI permits overlapping shared
buffers, but with the new FF-A based ABI each physical page may only
be registered once.

As the key and blob buffer are allocated adjancently, there is no need
for redundant register shared memory invocation. Also, it is incompatibile
with FF-A based ABI limitation. So refactor register shared memory
implementation to use only single invocation to register both key and blob
buffers.

[jarkko: Added cc to stable.]
Cc: stable@vger.kernel.org # v5.16+
Fixes: 4615e5a34b95 ("optee: add FF-A support")
Reported-by: Jens Wiklander <jens.wiklander@linaro.org>
Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Tested-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

pmdomain: bcm: bcm2835-power: check if the ASB register is equal to enable

commit 2e75396f1df61e1f1d26d0d703fc7292c4ae4371 upstream.

The commit c494a447c14e ("soc: bcm: bcm2835-power: Refactor ASB control")
refactored the ASB control by using a general function to handle both
the enable and disable. But this patch introduced a subtle regression:
we need to check if !!(readl(base + reg) & ASB_ACK) == enable, not just
check if (readl(base + reg) & ASB_ACK) == true.

Currently, this is causing an invalid register state in V3D when
unloading and loading the driver, because `bcm2835_asb_disable()` will
return -ETIMEDOUT and `bcm2835_asb_power_off()` will fail to disable the
ASB slave for V3D.

Fixes: c494a447c14e ("soc: bcm: bcm2835-power: Refactor ASB control")
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Stefan Wahren <stefan.wahren@i2se.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231024101251.6357-2-mcanal@igalia.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sched/core: Fix RQCF_ACT_SKIP leak

commit 5ebde09d91707a4a9bec1e3d213e3c12ffde348f upstream.

Igor Raits and Bagas Sanjaya report a RQCF_ACT_SKIP leak warning.

This warning may be triggered in the following situations:

    CPU0                                      CPU1

__schedule()
  *rq->clock_update_flags <<= 1;*   unregister_fair_sched_group()
  pick_next_task_fair+0x4a/0x410      destroy_cfs_bandwidth()
    newidle_balance+0x115/0x3e0       for_each_possible_cpu(i) *i=0*
      rq_unpin_lock(this_rq, rf)      __cfsb_csd_unthrottle()
      raw_spin_rq_unlock(this_rq)
                                      rq_lock(*CPU0_rq*, &rf)
                                      rq_clock_start_loop_update()
                                      rq->clock_update_flags & RQCF_ACT_SKIP <--
      raw_spin_rq_lock(this_rq)

The purpose of RQCF_ACT_SKIP is to skip the update rq clock,
but the update is very early in __schedule(), but we clear
RQCF_*_SKIP very late, causing it to span that gap above
and triggering this warning.

In __schedule() we can clear the RQCF_*_SKIP flag immediately
after update_rq_clock() to avoid this RQCF_ACT_SKIP leak warning.
And set rq->clock_update_flags to RQCF_UPDATED to avoid
rq->clock_update_flags < RQCF_ACT_SKIP warning that may be triggered later.

Fixes: ebb83d84e49b ("sched/core: Avoid multiple calling update_rq_clock() in __cfsb_csd_unthrottle()")
Closes: https://lore.kernel.org/all/20230913082424.73252-1-jiahao.os@bytedance.com
Reported-by: Igor Raits <igor.raits@gmail.com>
Reported-by: Bagas Sanjaya <bagasdotme@gmail.com>
Suggested-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Hao Jia <jiahao.os@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/a5dd536d-041a-2ce9-f4b7-64d8d85c86dc@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

genirq/generic_chip: Make irq_remove_generic_chip() irqdomain aware

commit 5e7afb2eb7b2a7c81e9f608cbdf74a07606fd1b5 upstream.

irq_remove_generic_chip() calculates the Linux interrupt number for removing the
handler and interrupt chip based on gc::irq_base as a linear function of
the bit positions of set bits in the @msk argument.

When the generic chip is present in an irq domain, i.e. created with a call
to irq_alloc_domain_generic_chips(), gc::irq_base contains not the base
Linux interrupt number.  It contains the base hardware interrupt for this
chip. It is set to 0 for the first chip in the domain, 0 + N for the next
chip, where $N is the number of hardware interrupts per chip.

That means the Linux interrupt number cannot be calculated based on
gc::irq_base for irqdomain based chips without a domain map lookup, which
is currently missing.

Rework the code to take the irqdomain case into account and calculate the
Linux interrupt number by a irqdomain lookup of the domain specific
hardware interrupt number.

[ tglx: Massage changelog. Reshuffle the logic and add a proper comment. ]

Fixes: cfefd21e693d ("genirq: Add chip suspend and resume callbacks")
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231024150335.322282-1-herve.codina@bootlin.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: meson-gx: Remove setting of CMD_CFG_ERROR

commit 57925e16c9f7d18012bcf45bfa658f92c087981a upstream.

For the t7 and older SoC families, the CMD_CFG_ERROR has no effect.
Starting from SoC family C3, setting this bit without SG LINK data
address will cause the controller to generate an IRQ and stop working.

To fix it, don't set the bit CMD_CFG_ERROR anymore.

Fixes: 18f92bc02f17 ("mmc: meson-gx: make sure the descriptor is stopped on errors")
Signed-off-by: Rong Chen <rong.chen@amlogic.com>
Reviewed-by: Jerome Brunet <jbrunet@baylibre.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231026073156.2868310-1-rong.chen@amlogic.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath12k: fix dfs-radar and temperature event locking

commit 69bd216e049349886405b1c87a55dce3d35d1ba7 upstream.

The ath12k active pdevs are protected by RCU but the DFS-radar and
temperature event handling code calling ath12k_mac_get_ar_by_pdev_id()
was not marked as a read-side critical section.

Mark the code in question as RCU read-side critical sections to avoid
any potential use-after-free issues.

Note that the temperature event handler looks like a place holder
currently but would still trigger an RCU lockdep splat.

Compile tested only.

Fixes: d889913205cf ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Cc: stable@vger.kernel.org # v6.2
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019113650.9060-2-johan+linaro@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath12k: fix htt mlo-offset event locking

commit 6afc57ea315e0f660b1f870a681737bb7b71faef upstream.

The ath12k active pdevs are protected by RCU but the htt mlo-offset
event handling code calling ath12k_mac_get_ar_by_pdev_id() was not
marked as a read-side critical section.

Mark the code in question as an RCU read-side critical section to avoid
any potential use-after-free issues.

Compile tested only.

Fixes: d889913205cf ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Cc: stable@vger.kernel.org # v6.2
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019113650.9060-3-johan+linaro@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath11k: fix gtk offload status event locking

commit 1dea3c0720a146bd7193969f2847ccfed5be2221 upstream.

The ath11k active pdevs are protected by RCU but the gtk offload status
event handling code calling ath11k_mac_get_arvif_by_vdev_id() was not
marked as a read-side critical section.

Mark the code in question as an RCU read-side critical section to avoid
any potential use-after-free issues.

Compile tested only.

Fixes: a16d9b50cfba ("ath11k: support GTK rekey offload")
Cc: stable@vger.kernel.org # 5.18
Cc: Carl Huang <quic_cjhuang@quicinc.com>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019155342.31631-1-johan+linaro@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath11k: fix htt pktlog locking

commit 3f77c7d605b29df277d77e9ee75d96e7ad145d2d upstream.

The ath11k active pdevs are protected by RCU but the htt pktlog handling
code calling ath11k_mac_get_ar_by_pdev_id() was not marked as a
read-side critical section.

Mark the code in question as an RCU read-side critical section to avoid
any potential use-after-free issues.

Compile tested only.

Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Cc: stable@vger.kernel.org # 5.6
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019112521.2071-1-johan+linaro@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath11k: fix dfs radar event locking

commit 3b6c14833165f689cc5928574ebafe52bbce5f1e upstream.

The ath11k active pdevs are protected by RCU but the DFS radar event
handling code calling ath11k_mac_get_ar_by_pdev_id() was not marked as a
read-side critical section.

Mark the code in question as an RCU read-side critical section to avoid
any potential use-after-free issues.

Compile tested only.

Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Cc: stable@vger.kernel.org # 5.6
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019153115.26401-3-johan+linaro@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath11k: fix temperature event locking

commit 1a5352a81b4720ba43d9c899974e3bddf7ce0ce8 upstream.

The ath11k active pdevs are protected by RCU but the temperature event
handling code calling ath11k_mac_get_ar_by_pdev_id() was not marked as a
read-side critical section as reported by RCU lockdep:

=============================
WARNING: suspicious RCU usage
6.6.0-rc6 #7 Not tainted
-----------------------------
drivers/net/wireless/ath/ath11k/mac.c:638 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

rcu_scheduler_active = 2, debug_locks = 1
no locks held by swapper/0/0.
...
Call trace:
...
 lockdep_rcu_suspicious+0x16c/0x22c
 ath11k_mac_get_ar_by_pdev_id+0x194/0x1b0 [ath11k]
 ath11k_wmi_tlv_op_rx+0xa84/0x2c1c [ath11k]
 ath11k_htc_rx_completion_handler+0x388/0x510 [ath11k]

Mark the code in question as an RCU read-side critical section to avoid
any potential use-after-free issues.

Tested-on: WCN6855 hw2.1 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.23

Fixes: a41d10348b01 ("ath11k: add thermal sensor device support")
Cc: stable@vger.kernel.org # 5.7
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20231019153115.26401-2-johan+linaro@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

regmap: Ensure range selector registers are updated after cache sync

commit 0ec7731655de196bc1e4af99e495b38778109d22 upstream.

When we sync the register cache we do so with the cache bypassed in order
to avoid overhead from writing the synced values back into the cache. If
the regmap has ranges and the selector register for those ranges is in a
register which is cached this has the unfortunate side effect of meaning
that the physical and cached copies of the selector register can be out of
sync after a cache sync. The cache will have whatever the selector was when
the sync started and the hardware will have the selector for the register
that was synced last.

Fix this by rewriting all cached selector registers after every sync,
ensuring that the hardware and cache have the same content. This will
result in extra writes that wouldn't otherwise be needed but is simple
so hopefully robust. We don't read from the hardware since not all
devices have physical read support.

Given that nobody noticed this until now it is likely that we are rarely if
ever hitting this case.

Reported-by: Hector Martin <marcan@marcan.st>
Cc: stable@vger.kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20231026-regmap-fix-selector-sync-v1-1-633ded82770d@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ACPI: resource: Do IRQ override on TongFang GMxXGxx

commit 0da9eccde3270b832c059ad618bf66e510c75d33 upstream.

The TongFang GMxXGxx/TUXEDO Stellaris/Pollaris Gen5 needs IRQ overriding
for the keyboard to work.

Adding an entry for this laptop to the override_table makes the internal
keyboard functional.

Signed-off-by: Werner Sembach <wse@tuxedocomputers.com>
Cc: All applicable <stable@vger.kernel.org>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: Add nop instructions after TLB inserts

commit ad4aa06e1d92b06ed56c7240252927bd60632efe upstream.

An excerpt from the PA8800 ERS states:

* The PA8800 violates the seven instruction pipeline rule when performing
  TLB inserts or PxTLBE instructions with the PSW C bit on. The instruction
  will take effect by the 12th instruction after the insert or purge.

I believe we have a problem with handling TLB misses. We don't fill
the pipeline following TLB inserts. As a result, we likely fault again
after returning from the interruption.

The above statement indicates that we need at least seven instructions
after the insert on pre PA8800 processors and we need 12 instructions
on PA8800/PA8900 processors.

Here we add macros and code to provide the required number instructions
after a TLB insert.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Suggested-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/sysfs: check error from damon_sysfs_update_target()

commit b4936b544b08ed44949055b92bd25f77759ebafc upstream.

Patch series "mm/damon/sysfs: fix unhandled return values".

Some of DAMON sysfs interface code is not handling return values from some
functions.  As a result, confusing user input handling or NULL-dereference
is possible.  Check those properly.

This patch (of 3):

damon_sysfs_update_target() returns error code for failures, but its
caller, damon_sysfs_set_targets() is ignoring that.  The update function
seems making no critical change in case of such failures, but the behavior
will look like DAMON sysfs is silently ignoring or only partially
accepting the user input.  Fix it.

Link: https://lkml.kernel.org/r/20231106233408.51159-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20231106233408.51159-2-sj@kernel.org
Fixes: 19467a950b49 ("mm/damon/sysfs: remove requested targets when online-commit inputs")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [5.19+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/core.c: avoid unintentional filtering out of schemes

commit 13b2a4b22e98ff80b888a160a2acd92d81b05925 upstream.

The function '__damos_filter_out()' causes DAMON to always filter out
schemes whose filter type is anon or memcg if its matching value is set
to false.

This commit addresses the issue by ensuring that '__damos_filter_out()'
no longer applies to filters whose type is 'anon' or 'memcg'.

Link: https://lkml.kernel.org/r/1699594629-3816-1-git-send-email-hyeongtak.ji@gmail.com
Fixes: ab9bda001b681 ("mm/damon/core: introduce address range type damos filter")
Signed-off-by: Hyeongtak Ji <hyeongtak.ji@sk.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/sysfs-schemes: handle tried regions sysfs directory allocation failure

commit 84055688b6bc075c92a88e2d6c3ad26ab93919f9 upstream.

DAMOS tried regions sysfs directory allocation function
(damon_sysfs_scheme_regions_alloc()) is not handling the memory allocation
failure.  In the case, the code will dereference NULL pointer.  Handle the
failure to avoid such invalid access.

Link: https://lkml.kernel.org/r/20231106233408.51159-3-sj@kernel.org
Fixes: 9277d0367ba1 ("mm/damon/sysfs-schemes: implement scheme region directory")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [6.2+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/sysfs-schemes: handle tried region directory allocation failure

commit ae636ae2bbfd9279f5681dbf320d1da817e52b68 upstream.

DAMON sysfs interface's before_damos_apply callback
(damon_sysfs_before_damos_apply()), which creates the DAMOS tried regions
for each DAMOS action applied region, is not handling the allocation
failure for the sysfs directory data.  As a result, NULL pointer
derefeence is possible.  Fix it by handling the case.

Link: https://lkml.kernel.org/r/20231106233408.51159-4-sj@kernel.org
Fixes: f1d13cacabe1 ("mm/damon/sysfs: implement DAMOS tried regions update command")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [6.2+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/core: avoid divide-by-zero during monitoring results update

commit d35963bfb05877455228ecec6b194f624489f96a upstream.

When monitoring attributes are changed, DAMON updates access rate of the
monitoring results accordingly.  For that, it divides some values by the
maximum nr_accesses.  However, due to the type of the related variables,
simple division-based calculation of the divisor can return zero.  As a
result, divide-by-zero is possible.  Fix it by using
damon_max_nr_accesses(), which handles the case.

Link: https://lkml.kernel.org/r/20231019194924.100347-3-sj@kernel.org
Fixes: 2f5bef5a590b ("mm/damon/core: update monitoring results for new monitoring attributes")
Signed-off-by: SeongJae Park <sj@kernel.org>
Reported-by: Jakub Acs <acsjakub@amazon.de>
Cc: <stable@vger.kernel.org> [6.3+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon: implement a function for max nr_accesses safe calculation

commit 35f5d94187a6a3a8df2cba54beccca1c2379edb8 upstream.

Patch series "avoid divide-by-zero due to max_nr_accesses overflow".

The maximum nr_accesses of given DAMON context can be calculated by
dividing the aggregation interval by the sampling interval.  Some logics
in DAMON uses the maximum nr_accesses as a divisor.  Hence, the value
shouldn't be zero.  Such case is avoided since DAMON avoids setting the
agregation interval as samller than the sampling interval.  However, since
nr_accesses is unsigned int while the intervals are unsigned long, the
maximum nr_accesses could be zero while casting.

Avoid the divide-by-zero by implementing a function that handles the
corner case (first patch), and replaces the vulnerable direct max
nr_accesses calculations (remaining patches).

Note that the patches for the replacements are divided for broken commits,
to make backporting on required tres easier.  Especially, the last patch
is for a patch that not yet merged into the mainline but in mm tree.

This patch (of 4):

The maximum nr_accesses of given DAMON context can be calculated by
dividing the aggregation interval by the sampling interval.  Some logics
in DAMON uses the maximum nr_accesses as a divisor.  Hence, the value
shouldn't be zero.  Such case is avoided since DAMON avoids setting the
agregation interval as samller than the sampling interval.  However, since
nr_accesses is unsigned int while the intervals are unsigned long, the
maximum nr_accesses could be zero while casting.  Implement a function
that handles the corner case.

Note that this commit is not fixing the real issue since this is only
introducing the safe function that will replaces the problematic
divisions.  The replacements will be made by followup commits, to make
backporting on stable series easier.

Link: https://lkml.kernel.org/r/20231019194924.100347-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20231019194924.100347-2-sj@kernel.org
Fixes: 198f0f4c58b9 ("mm/damon/vaddr,paddr: support pageout prioritization")
Signed-off-by: SeongJae Park <sj@kernel.org>
Reported-by: Jakub Acs <acsjakub@amazon.de>
Cc: <stable@vger.kernel.org> [5.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/ops-common: avoid divide-by-zero during region hotness calculation

commit 3bafc47d3c4a2fc4d3b382aeb3c087f8fc84d9fd upstream.

When calculating the hotness of each region for the under-quota regions
prioritization, DAMON divides some values by the maximum nr_accesses.
However, due to the type of the related variables, simple division-based
calculation of the divisor can return zero.  As a result, divide-by-zero
is possible.  Fix it by using damon_max_nr_accesses(), which handles the
case.

Link: https://lkml.kernel.org/r/20231019194924.100347-4-sj@kernel.org
Fixes: 198f0f4c58b9 ("mm/damon/vaddr,paddr: support pageout prioritization")
Signed-off-by: SeongJae Park <sj@kernel.org>
Reported-by: Jakub Acs <acsjakub@amazon.de>
Cc: <stable@vger.kernel.org> [5.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/lru_sort: avoid divide-by-zero in hot threshold calculation

commit 44063f125af4bb4efd1d500d8091fa33a98af325 upstream.

When calculating the hotness threshold for lru_prio scheme of
DAMON_LRU_SORT, the module divides some values by the maximum nr_accesses.
However, due to the type of the related variables, simple division-based
calculation of the divisor can return zero.  As a result, divide-by-zero
is possible.  Fix it by using damon_max_nr_accesses(), which handles the
case.

Link: https://lkml.kernel.org/r/20231019194924.100347-5-sj@kernel.org
Fixes: 40e983cca927 ("mm/damon: introduce DAMON-based LRU-lists Sorting")
Signed-off-by: SeongJae Park <sj@kernel.org>
Reported-by: Jakub Acs <acsjakub@amazon.de>
Cc: <stable@vger.kernel.org> [6.0+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm crypt: account large pages in cc->n_allocated_pages

commit 9793c269da6cd339757de6ba5b2c8681b54c99af upstream.

The commit 5054e778fcd9c ("dm crypt: allocate compound pages if
possible") changed dm-crypt to use compound pages to improve
performance. Unfortunately, there was an oversight: the allocation of
compound pages was not accounted at all. Normal pages are accounted in
a percpu counter cc->n_allocated_pages and dm-crypt is limited to
allocate at most 2% of memory. Because compound pages were not
accounted at all, dm-crypt could allocate memory over the 2% limit.

Fix this by adding the accounting of compound pages, so that memory
consumption of dm-crypt is properly limited.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Fixes: 5054e778fcd9c ("dm crypt: allocate compound pages if possible")
Cc: stable@vger.kernel.org # v6.5+
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fbdev: stifb: Make the STI next font pointer a 32-bit signed offset

commit 8a32aa17c1cd48df1ddaa78e45abcb8c7a2220d6 upstream.

The pointer to the next STI font is actually a signed 32-bit
offset. With this change the 64-bit kernel will correctly subract
the (signed 32-bit) offset instead of adding a (unsigned 32-bit)
offset. It has no effect on 32-bit kernels.

This fixes the stifb driver with a 64-bit kernel on qemu.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iommufd: Fix missing update of domains_itree after splitting iopt_area

commit e7250ab7ca4998fe026f2149805b03e09dc32498 upstream.

In iopt_area_split(), if the original iopt_area has filled a domain and is
linked to domains_itree, pages_nodes have to be properly
reinserted. Otherwise the domains_itree becomes corrupted and we will UAF.

Fixes: 51fe6141f0f6 ("iommufd: Data structure to provide IOVA to PFN mapping")
Link: https://lore.kernel.org/r/20231027162941.2864615-2-den@valinux.co.jp
Cc: stable@vger.kernel.org
Signed-off-by: Koichiro Den <den@valinux.co.jp>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

watchdog: move softlockup_panic back to early_param

commit 8b793bcda61f6c3ed4f5b2ded7530ef6749580cb upstream.

Setting softlockup_panic from do_sysctl_args() causes it to take effect
later in boot.  The lockup detector is enabled before SMP is brought
online, but do_sysctl_args runs afterwards.  If a user wants to set
softlockup_panic on boot and have it trigger should a softlockup occur
during onlining of the non-boot processors, they could do this prior to
commit f117955a2255 ("kernel/watchdog.c: convert {soft/hard}lockup boot
parameters to sysctl aliases").  However, after this commit the value
of softlockup_panic is set too late to be of help for this type of
problem.  Restore the prior behavior.

Signed-off-by: Krister Johansen <kjlx@templeofstupid.com>
Cc: stable@vger.kernel.org
Fixes: f117955a2255 ("kernel/watchdog.c: convert {soft/hard}lockup boot parameters to sysctl aliases")
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/sysfs: update monitoring target regions for online input commit

commit 9732336006764e2ee61225387e3c70eae9139035 upstream.

When user input is committed online, DAMON sysfs interface is ignoring the
user input for the monitoring target regions.  Such request is valid and
useful for fixed monitoring target regions-based monitoring ops like
'paddr' or 'fvaddr'.

Update the region boundaries as user specified, too.  Note that the
monitoring results of the regions that overlap between the latest
monitoring target regions and the new target regions are preserved.

Treat empty monitoring target regions user request as a request to just
make no change to the monitoring target regions.  Otherwise, users should
set the monitoring target regions same to current one for every online
input commit, and it could be challenging for dynamic monitoring target
regions update DAMON ops like 'vaddr'.  If the user really need to remove
all monitoring target regions, they can simply remove the target and then
create the target again with empty target regions.

Link: https://lkml.kernel.org/r/20231031170131.46972-1-sj@kernel.org
Fixes: da87878010e5 ("mm/damon/sysfs: support online inputs update")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [5.19+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/sysfs: remove requested targets when online-commit inputs

commit 19467a950b49432a84bf6dbadbbb17bdf89418b7 upstream.

damon_sysfs_set_targets(), which updates the targets of the context for
online commitment, do not remove targets that removed from the
corresponding sysfs files.  As a result, more than intended targets of the
context can exist and hence consume memory and monitoring CPU resource
more than expected.

Fix it by removing all targets of the context and fill up again using the
user input.  This could cause unnecessary memory dealloc and realloc
operations, but this is not a hot code path.  Also, note that damon_target
is stateless, and hence no data is lost.

[sj@kernel.org: fix unnecessary monitoring results removal]
Link: https://lkml.kernel.org/r/20231028213353.45397-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20231022210735.46409-2-sj@kernel.org
Fixes: da87878010e5 ("mm/damon/sysfs: support online inputs update")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Cc: <stable@vger.kernel.org> [5.19.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

PCI/sysfs: Protect driver's D3cold preference from user space

commit 70b70a4307cccebe91388337b1c85735ce4de6ff upstream.

struct pci_dev contains two flags which govern whether the device may
suspend to D3cold:

* no_d3cold provides an opt-out for drivers (e.g. if a device is known
  to not wake from D3cold)

* d3cold_allowed provides an opt-out for user space (default is true,
  user space may set to false)

Since commit 9d26d3a8f1b0 ("PCI: Put PCIe ports into D3 during suspend"),
the user space setting overwrites the driver setting.  Essentially user
space is trusted to know better than the driver whether D3cold is
working.

That feels unsafe and wrong.  Assume that the change was introduced
inadvertently and do not overwrite no_d3cold when d3cold_allowed is
modified.  Instead, consider d3cold_allowed in addition to no_d3cold
when choosing a suspend state for the device.

That way, user space may opt out of D3cold if the driver hasn't, but it
may no longer force an opt in if the driver has opted out.

Fixes: 9d26d3a8f1b0 ("PCI: Put PCIe ports into D3 during suspend")
Link: https://lore.kernel.org/r/b8a7f4af2b73f6b506ad8ddee59d747cbf834606.1695025365.git.lukas@wunner.de
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Cc: stable@vger.kernel.org # v4.8+
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

hvc/xen: fix event channel handling for secondary consoles

commit ef5dd8ec88ac11e8e353164407d55b73c988b369 upstream.

The xencons_connect_backend() function allocates a local interdomain
event channel with xenbus_alloc_evtchn(), then calls
bind_interdomain_evtchn_to_irq_lateeoi() to bind to that port# on the
*remote* domain.

That doesn't work very well:

(qemu) device_add xen-console,id=con1,chardev=pty0
[   44.323872] xenconsole console-1: 2 xenbus_dev_probe on device/console/1
[   44.323995] xenconsole: probe of console-1 failed with error -2

Fix it to use bind_evtchn_to_irq_lateeoi(), which does the right thing
by just binding that *local* event channel to an irq. The backend will
do the interdomain binding.

This didn't affect the primary console because the setup for that is
special — the toolstack allocates the guest event channel and the guest
discovers it with HVMOP_get_param.

Fixes: fe415186b43d ("xen/console: harden hvc_xen against event channel storms")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Juergen Gross <jgross@suse.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231020161529.355083-2-dwmw2@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

hvc/xen: fix error path in xen_hvc_init() to always register frontend driver

commit 2704c9a5593f4a47620c12dad78838ca62b52f48 upstream.

The xen_hvc_init() function should always register the frontend driver,
even when there's no primary console — as there may be secondary consoles.
(Qemu can always add secondary consoles, but only the toolstack can add
the primary because it's special.)

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Juergen Gross <jgross@suse.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231020161529.355083-3-dwmw2@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

hvc/xen: fix console unplug

commit a30badfd7c13fc8763a9e10c5a12ba7f81515a55 upstream.

On unplug of a Xen console, xencons_disconnect_backend() unconditionally
calls free_irq() via unbind_from_irqhandler(), causing a warning of
freeing an already-free IRQ:

(qemu) device_del con1
[   32.050919] ------------[ cut here ]------------
[   32.050942] Trying to free already-free IRQ 33
[   32.050990] WARNING: CPU: 0 PID: 51 at kernel/irq/manage.c:1895 __free_irq+0x1d4/0x330

It should be using evtchn_put() to tear down the event channel binding,
and let the Linux IRQ side of it be handled by notifier_del_irq() through
the HVC code.

On which topic... xencons_disconnect_backend() should call hvc_remove()
*first*, rather than tearing down the event channel and grant mapping
while they are in use. And then the IRQ is guaranteed to be freed by
the time it's torn down by evtchn_put().

Since evtchn_put() also closes the actual event channel, avoid calling
xenbus_free_evtchn() except in the failure path where the IRQ was not
successfully set up.

However, calling hvc_remove() at the start of xencons_disconnect_backend()
still isn't early enough. An unplug request is indicated by the backend
setting its state to XenbusStateClosing, which triggers a notification
to xencons_backend_changed(), which... does nothing except set its own
frontend state directly to XenbusStateClosed without *actually* tearing
down the HVC device or, you know, making sure it isn't actively in use.

So the backend sees the guest frontend set its state to XenbusStateClosed
and stops servicing the interrupt... and the guest spins for ever in the
domU_write_console() function waiting for the ring to drain.

Fix that one by calling hvc_remove() from xencons_backend_changed() before
signalling to the backend that it's OK to proceed with the removal.

Tested with 'dd if=/dev/zero of=/dev/hvc1' while telling Qemu to remove
the console device.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231020161529.355083-4-dwmw2@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

acpi/processor: sanitize _OSC/_PDC capabilities for Xen dom0

commit bfa993b355d33a438a746523e7129391c8664e8a upstream.

The Processor capability bits notify ACPI of the OS capabilities, and
so ACPI can adjust the return of other Processor methods taking the OS
capabilities into account.

When Linux is running as a Xen dom0, the hypervisor is the entity
in charge of processor power management, and hence Xen needs to make
sure the capabilities reported by _OSC/_PDC match the capabilities of
the driver in Xen.

Introduce a small helper to sanitize the buffer when running as Xen
dom0.

When Xen supports HWP, this serves as the equivalent of commit
a21211672c9a ("ACPI / processor: Request native thermal interrupt
handling via _OSC") to avoid SMM crashes.  Xen will set bit
ACPI_PROC_CAP_COLLAB_PROC_PERF (bit 12) in the capability bits and the
_OSC/_PDC call will apply it.

[ jandryuk: Mention Xen HWP's need.  Support _OSC & _PDC ]
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: stable@vger.kernel.org
Signed-off-by: Jason Andryuk <jandryuk@gmail.com>
Reviewed-by: Michal Wilczynski <michal.wilczynski@intel.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20231108212517.72279-1-jandryuk@gmail.com
Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tty: serial: meson: fix hard LOCKUP on crtscts mode

commit 2a1d728f20edeee7f26dc307ed9df4e0d23947ab upstream.

There might be hard lockup if we set crtscts mode on port without RTS/CTS configured:

# stty -F /dev/ttyAML6 crtscts; echo 1 > /dev/ttyAML6; echo 2 > /dev/ttyAML6
[   95.890386] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
[   95.890857] rcu:     3-...0: (201 ticks this GP) idle=e33c/1/0x4000000000000000 softirq=5844/5846 fqs=4984
[   95.900212] rcu:     (detected by 2, t=21016 jiffies, g=7753, q=296 ncpus=4)
[   95.906972] Task dump for CPU 3:
[   95.910178] task:bash            state:R  running task     stack:0     pid:205   ppid:1      flags:0x00000202
[   95.920059] Call trace:
[   95.922485]  __switch_to+0xe4/0x168
[   95.925951]  0xffffff8003477508
[   95.974379] watchdog: Watchdog detected hard LOCKUP on cpu 3
[   95.974424] Modules linked in: 88x2cs(O) rtc_meson_vrtc

Possible solution would be to not allow to setup crtscts on such port.

Tested on S905X3 based board.

Fixes: ff7693d079e5 ("ARM: meson: serial: add MesonX SoC on-chip uart driver")
Cc: stable@vger.kernel.org
Signed-off-by: Pavel Krasavin <pkrasavin@imaqliq.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Dmitry Rokosov <ddrokosov@salutedevices.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tty/sysrq: replace smp_processor_id() with get_cpu()

commit dd976a97d15b47656991e185a94ef42a0fa5cfd4 upstream.

The smp_processor_id() shouldn't be called from preemptible code.
Instead use get_cpu() and put_cpu() which disables preemption in
addition to getting the processor id. Enable preemption back after
calling schedule_work() to make sure that the work gets scheduled on all
cores other than the current core. We want to avoid a scenario where
current core's stack trace is printed multiple times and one core's
stack trace isn't printed because of scheduling of current task.

This fixes the following bug:

[  119.143590] sysrq: Show backtrace of all active CPUs
[  119.143902] BUG: using smp_processor_id() in preemptible [00000000] code: bash/873
[  119.144586] caller is debug_smp_processor_id+0x20/0x30
[  119.144827] CPU: 6 PID: 873 Comm: bash Not tainted 5.10.124-dirty #3
[  119.144861] Hardware name: QEMU QEMU Virtual Machine, BIOS 2023.05-1 07/22/2023
[  119.145053] Call trace:
[  119.145093]  dump_backtrace+0x0/0x1a0
[  119.145122]  show_stack+0x18/0x70
[  119.145141]  dump_stack+0xc4/0x11c
[  119.145159]  check_preemption_disabled+0x100/0x110
[  119.145175]  debug_smp_processor_id+0x20/0x30
[  119.145195]  sysrq_handle_showallcpus+0x20/0xc0
[  119.145211]  __handle_sysrq+0x8c/0x1a0
[  119.145227]  write_sysrq_trigger+0x94/0x12c
[  119.145247]  proc_reg_write+0xa8/0xe4
[  119.145266]  vfs_write+0xec/0x280
[  119.145282]  ksys_write+0x6c/0x100
[  119.145298]  __arm64_sys_write+0x20/0x30
[  119.145315]  el0_svc_common.constprop.0+0x78/0x1e4
[  119.145332]  do_el0_svc+0x24/0x8c
[  119.145348]  el0_svc+0x10/0x20
[  119.145364]  el0_sync_handler+0x134/0x140
[  119.145381]  el0_sync+0x180/0x1c0

Cc: jirislaby@kernel.org
Cc: stable@vger.kernel.org
Fixes: 47cab6a722d4 ("debug lockups: Improve lockup detection, fix generic arch fallback")
Signed-off-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Link: https://lore.kernel.org/r/20231009162021.3607632-1-usama.anjum@collabora.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

proc: sysctl: prevent aliased sysctls from getting passed to init

commit 8001f49394e353f035306a45bcf504f06fca6355 upstream.

The code that checks for unknown boot options is unaware of the sysctl
alias facility, which maps bootparams to sysctl values.  If a user sets
an old value that has a valid alias, a message about an invalid
parameter will be printed during boot, and the parameter will get passed
to init.  Fix by checking for the existence of aliased parameters in the
unknown boot parameter code.  If an alias exists, don't return an error
or pass the value to init.

Signed-off-by: Krister Johansen <kjlx@templeofstupid.com>
Cc: stable@vger.kernel.org
Fixes: 0a477e1ae21b ("kernel/sysctl: support handling command line aliases")
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

audit: don't WARN_ON_ONCE(!current->mm) in audit_exe_compare()

commit 969d90ec212bae4b45bf9d21d7daa30aa6cf055e upstream.

eBPF can end up calling into the audit code from some odd places, and
some of these places don't have @current set properly so we end up
tripping the `WARN_ON_ONCE(!current->mm)` near the top of
`audit_exe_compare()`.  While the basic `!current->mm` check is good,
the `WARN_ON_ONCE()` results in some scary console messages so let's
drop that and just do the regular `!current->mm` check to avoid
problems.

Cc: <stable@vger.kernel.org>
Fixes: 47846d51348d ("audit: don't take task_lock() in audit_exe_compare() code path")
Reported-by: Artem Savkov <asavkov@redhat.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

audit: don't take task_lock() in audit_exe_compare() code path

commit 47846d51348dd62e5231a83be040981b17c955fa upstream.

The get_task_exe_file() function locks the given task with task_lock()
which when used inside audit_exe_compare() can cause deadlocks on
systems that generate audit records when the task_lock() is held. We
resolve this problem with two changes: ignoring those cases where the
task being audited is not the current task, and changing our approach
to obtaining the executable file struct to not require task_lock().

With the intent of the audit exe filter being to filter on audit events
generated by processes started by the specified executable, it makes
sense that we would only want to use the exe filter on audit records
associated with the currently executing process, e.g. @current.  If
we are asked to filter records using a non-@current task_struct we can
safely ignore the exe filter without negatively impacting the admin's
expectations for the exe filter.

Knowing that we only have to worry about filtering the currently
executing task in audit_exe_compare() we can do away with the
task_lock() and call get_mm_exe_file() with @current->mm directly.

Cc: <stable@vger.kernel.org>
Fixes: 5efc244346f9 ("audit: fix exe_file access in audit_exe_compare")
Reported-by: Andreas Steinmetz <anstein99@googlemail.com>
Reviewed-by: John Johansen <john.johanse@canonical.com>
Reviewed-by: Mateusz Guzik <mjguzik@gmail.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sched: psi: fix unprivileged polling against cgroups

commit 8b39d20eceeda6c4eb23df1497f9ed2fffdc8f69 upstream.

519fabc7aaba ("psi: remove 500ms min window size limitation for
triggers") breaks unprivileged psi polling on cgroups.

Historically, we had a privilege check for polling in the open() of a
pressure file in /proc, but were erroneously missing it for the open()
of cgroup pressure files.

When unprivileged polling was introduced in d82caa273565 ("sched/psi:
Allow unprivileged polling of N*2s period"), it needed to filter
privileges depending on the exact polling parameters, and as such
moved the CAP_SYS_RESOURCE check from the proc open() callback to
psi_trigger_create(). Both the proc files as well as cgroup files go
through this during write(). This implicitly added the missing check
for privileges required for HT polling for cgroups.

When 519fabc7aaba ("psi: remove 500ms min window size limitation for
triggers") followed right after to remove further restrictions on the
RT polling window, it incorrectly assumed the cgroup privilege check
was still missing and added it to the cgroup open(), mirroring what we
used to do for proc files in the past.

As a result, unprivileged poll requests that would be supported now
get rejected when opening the cgroup pressure file for writing.

Remove the cgroup open() check. psi_trigger_create() handles it.

Fixes: 519fabc7aaba ("psi: remove 500ms min window size limitation for triggers")
Reported-by: Luca Boccassi <bluca@debian.org>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Luca Boccassi <bluca@debian.org>
Acked-by: Suren Baghdasaryan <surenb@google.com>
Cc: stable@vger.kernel.org # 6.5+
Link: https://lore.kernel.org/r/20231026164114.2488682-1-hannes@cmpxchg.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: sdhci-pci-gli: GL9755: Mask the replay timer timeout of AER

commit 85dd3af64965c1c0eb7373b340a1b1f7773586b0 upstream.

Due to a flaw in the hardware design, the GL9755 replay timer frequently
times out when ASPM is enabled. As a result, the warning messages will
often appear in the system log when the system accesses the GL9755
PCI config. Therefore, the replay timer timeout must be masked.

Fixes: 36ed2fd32b2c ("mmc: sdhci-pci-gli: A workaround to allow GL9755 to enter ASPM L1.2")
Signed-off-by: Victor Shih <victor.shih@genesyslogic.com.tw>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Kai-Heng Feng <kai.heng.geng@canonical.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231107095741.8832-3-victorshihgli@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: Fix lapic timer interrupt lost after loading a snapshot.

commit 9cfec6d097c607e36199cf0cfbb8cf5acbd8e9b2 upstream.

When running android emulator (which is based on QEMU 2.12) on
certain Intel hosts with kernel version 6.3-rc1 or above, guest
will freeze after loading a snapshot. This is almost 100%
reproducible. By default, the android emulator will use snapshot
to speed up the next launching of the same android guest. So
this breaks the android emulator badly.

I tested QEMU 8.0.4 from Debian 12 with an Ubuntu 22.04 guest by
running command "loadvm" after "savevm". The same issue is
observed. At the same time, none of our AMD platforms is impacted.
More experiments show that loading the KVM module with
"enable_apicv=false" can workaround it.

The issue started to show up after commit 8e6ed96cdd50 ("KVM: x86:
fire timer when it is migrated and expired, and in oneshot mode").
However, as is pointed out by Sean Christopherson, it is introduced
by commit 967235d32032 ("KVM: vmx: clear pending interrupts on
KVM_SET_LAPIC"). commit 8e6ed96cdd50 ("KVM: x86: fire timer when
it is migrated and expired, and in oneshot mode") just makes it
easier to hit the issue.

Having both commits, the oneshot lapic timer gets fired immediately
inside the KVM_SET_LAPIC call when loading the snapshot. On Intel
platforms with APIC virtualization and posted interrupt processing,
this eventually leads to setting the corresponding PIR bit. However,
the whole PIR bits get cleared later in the same KVM_SET_LAPIC call
by apicv_post_state_restore. This leads to timer interrupt lost.

The fix is to move vmx_apicv_post_state_restore to the beginning of
the KVM_SET_LAPIC call and rename to vmx_apicv_pre_state_restore.
What vmx_apicv_post_state_restore does is actually clearing any
former apicv state and this behavior is more suitable to carry out
in the beginning.

Fixes: 967235d32032 ("KVM: vmx: clear pending interrupts on KVM_SET_LAPIC")
Cc: stable@vger.kernel.org
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Haitao Shan <hshan@google.com>
Link: https://lore.kernel.org/r/20230913000215.478387-1-hshan@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: Clear bit12 of ICR after APIC-write VM-exit

commit 629d3698f6958ee6f8131ea324af794f973b12ac upstream.

When IPI virtualization is enabled, a WARN is triggered if bit12 of ICR
MSR is set after APIC-write VM-exit. The reason is kvm_apic_send_ipi()
thinks the APIC_ICR_BUSY bit should be cleared because KVM has no delay,
but kvm_apic_write_nodecode() doesn't clear the APIC_ICR_BUSY bit.

Under the x2APIC section, regarding ICR, the SDM says:

  It remains readable only to aid in debugging; however, software should
  not assume the value returned by reading the ICR is the last written
  value.

I.e. the guest is allowed to set bit 12.  However, the SDM also gives KVM
free reign to do whatever it wants with the bit, so long as KVM's behavior
doesn't confuse userspace or break KVM's ABI.

Clear bit 12 so that it reads back as '0'. This approach is safer than
"do nothing" and is consistent with the case where IPI virtualization is
disabled or not supported, i.e.,

  handle_fastpath_set_x2apic_icr_irqoff() -> kvm_x2apic_icr_write()

Opportunistically replace the TODO with a comment calling out that eating
the write is likely faster than a conditional branch around the busy bit.

Link: https://lore.kernel.org/all/ZPj6iF0Q7iynn62p@google.com/
Fixes: 5413bcba7ed5 ("KVM: x86: Add support for vICR APIC-write VM-Exits in x2APIC mode")
Cc: stable@vger.kernel.org
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Tested-by: Yi Lai <yi1.lai@intel.com>
Reviewed-by: Chao Gao <chao.gao@intel.com>
Link: https://lore.kernel.org/r/20230914055504.151365-1-tao1.su@linux.intel.com
[sean: tweak changelog, replace TODO with comment, drop local "val"]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: Ignore MSR_AMD64_TW_CFG access

commit 2770d4722036d6bd24bcb78e9cd7f6e572077d03 upstream.

Hyper-V enabled Windows Server 2022 KVM VM cannot be started on Zen1 Ryzen
since it crashes at boot with SYSTEM_THREAD_EXCEPTION_NOT_HANDLED +
STATUS_PRIVILEGED_INSTRUCTION (in other words, because of an unexpected #GP
in the guest kernel).

This is because Windows tries to set bit 8 in MSR_AMD64_TW_CFG and can't
handle receiving a #GP when doing so.

Give this MSR the same treatment that commit 2e32b7190641
("x86, kvm: Add MSR_AMD64_BU_CFG2 to the list of ignored MSRs") gave
MSR_AMD64_BU_CFG2 under justification that this MSR is baremetal-relevant
only.
Although apparently it was then needed for Linux guests, not Windows as in
this case.

With this change, the aforementioned guest setup is able to finish booting
successfully.

This issue can be reproduced either on a Summit Ridge Ryzen (with
just "-cpu host") or on a Naples EPYC (with "-cpu host,stepping=1" since
EPYC is ordinarily stepping 2).

Alternatively, userspace could solve the problem by using MSR filters, but
forcing every userspace to define a filter isn't very friendly and doesn't
add much, if any, value.  The only potential hiccup is if one of these
"baremetal-only" MSRs ever requires actual emulation and/or has F/M/S
specific behavior.  But if that happens, then KVM can still punt *that*
handling to userspace since userspace MSR filters "win" over KVM's default
handling.

Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/1ce85d9c7c9e9632393816cf19c902e0a3f411f1.1697731406.git.maciej.szmigiero@oracle.com
[sean: call out MSR filtering alternative]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: hyper-v: Don't auto-enable stimer on write from user-space

commit d6800af51c76b6dae20e6023bbdc9b3da3ab5121 upstream.

Don't apply the stimer's counter side effects when modifying its
value from user-space, as this may trigger spurious interrupts.

For example:
 - The stimer is configured in auto-enable mode.
 - The stimer's count is set and the timer enabled.
 - The stimer expires, an interrupt is injected.
 - The VM is live migrated.
 - The stimer config and count are deserialized, auto-enable is ON, the
   stimer is re-enabled.
 - The stimer expires right away, and injects an unwarranted interrupt.

Cc: stable@vger.kernel.org
Fixes: 1f4b34f825e8 ("kvm/x86: Hyper-V SynIC timers")
Signed-off-by: Nicolas Saenz Julienne <nsaenz@amazon.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Link: https://lore.kernel.org/r/20231017155101.40677-1-nsaenz@amazon.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu/hygon: Fix the CPU topology evaluation for real

commit ee545b94d39a00c93dc98b1dbcbcf731d2eadeb4 upstream.

Hygon processors with a model ID > 3 have CPUID leaf 0xB correctly
populated and don't need the fixed package ID shift workaround. The fixup
is also incorrect when running in a guest.

Fixes: e0ceeae708ce ("x86/CPU/hygon: Fix phys_proc_id calculation logic for multi-die processors")
Signed-off-by: Pu Wen <puwen@hygon.cn>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/tencent_594804A808BD93A4EBF50A994F228E3A7F07@qq.com
Link: https://lore.kernel.org/r/20230814085112.089607918@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/apic/msi: Fix misconfigured non-maskable MSI quirk

commit b56ebe7c896dc78b5865ec2c4b1dae3c93537517 upstream.

commit ef8dd01538ea ("genirq/msi: Make interrupt allocation less
convoluted"), reworked the code so that the x86 specific quirk for affinity
setting of non-maskable PCI/MSI interrupts is not longer activated if
necessary.

This could be solved by restoring the original logic in the core MSI code,
but after a deeper analysis it turned out that the quirk flag is not
required at all.

The quirk is only required when the PCI/MSI device cannot mask the MSI
interrupts, which in turn also prevents reservation mode from being enabled
for the affected interrupt.

This allows ot remove the NOMASK quirk bit completely as msi_set_affinity()
can instead check whether reservation mode is enabled for the interrupt,
which gives exactly the same answer.

Even in the momentary non-existing case that the reservation mode would be
not set for a maskable MSI interrupt this would not cause any harm as it
just would cause msi_set_affinity() to go needlessly through the
functionaly equivalent slow path, which works perfectly fine with maskable
interrupts as well.

Rework msi_set_affinity() to query the reservation mode and remove all
NOMASK quirk logic from the core code.

[ tglx: Massaged changelog ]

Fixes: ef8dd01538ea ("genirq/msi: Make interrupt allocation less convoluted")
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Koichiro Den <den@valinux.co.jp>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20231026032036.2462428-1-den@valinux.co.jp
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/PCI: Avoid PME from D3hot/D3cold for AMD Rembrandt and Phoenix USB4

commit 7d08f21f8c6307cb05cabb8d86e90ff6ccba57e9 upstream.

Iain reports that USB devices can't be used to wake a Lenovo Z13 from
suspend.  This occurs because on some AMD platforms, even though the Root
Ports advertise PME_Support for D3hot and D3cold, wakeup events from
devices on a USB4 controller don't result in wakeup interrupts from the
Root Port when amd-pmc has put the platform in a hardware sleep state.

If amd-pmc will be involved in the suspend, remove D3hot and D3cold from
the PME_Support mask of Root Ports above USB4 controllers so we avoid those
states if we need wakeups.

Restore D3 support at resume so that it can be used by runtime suspend.

This affects both AMD Rembrandt and Phoenix SoCs.

"pm_suspend_target_state == PM_SUSPEND_ON" means we're doing runtime
suspend, and amd-pmc will not be involved.  In that case PMEs work as
advertised in D3hot/D3cold, so we don't need to do anything.

Note that amd-pmc is technically optional, and there's no need for this
quirk if it's not present, but we assume it's always present because power
consumption is so high without it.

Fixes: 9d26d3a8f1b0 ("PCI: Put PCIe ports into D3 during suspend")
Link: https://lore.kernel.org/r/20231004144959.158840-1-mario.limonciello@amd.com
Reported-by: Iain Lane <iain@orangesquash.org.uk>
Closes: https://forums.lenovo.com/t5/Ubuntu/Z13-can-t-resume-from-suspend-with-external-USB-keyboard/m-p/5217121
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
[bhelgaas: commit log, move to arch/x86/pci/fixup.c, add #includes]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

crypto: x86/sha - load modules based on CPU features

commit 1c43c0f1f84aa59dfc98ce66f0a67b2922aa7f9d upstream.

x86 optimized crypto modules are built as modules rather than build-in and
they are not loaded when the crypto API is initialized, resulting in the
generic builtin module (sha1-generic) being used instead.

It was discovered when creating a sha1/sha256 checksum of a 2Gb file by
using kcapi-tools because it would take significantly longer than creating
a sha512 checksum of the same file. trace-cmd showed that for sha1/256 the
generic module was used, whereas for sha512 the optimized module was used
instead.

Add module aliases() for these x86 optimized crypto modules based on CPU
feature bits so udev gets a chance to load them later in the boot
process. This resulted in ~3x decrease in the real-time execution of
kcapi-dsg.

Fix is inspired from commit
aa031b8f702e ("crypto: x86/sha512 - load based on CPU features")
where a similar fix was done for sha512.

Cc: stable@vger.kernel.org # 5.15+
Suggested-by: Dimitri John Ledkov <dimitri.ledkov@canonical.com>
Suggested-by: Julian Andres Klode <julian.klode@canonical.com>
Signed-off-by: Roxana Nicolescu <roxana.nicolescu@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/shstk: Delay signal entry SSP write until after user accesses

commit 31255e072b2e91f97645d792d25b2db744186dd1 upstream.

When a signal is being delivered, the kernel needs to make accesses to
userspace. These accesses could encounter an access error, in which case
the signal delivery itself will trigger a segfault. Usually this would
result in the kernel killing the process. But in the case of a SEGV signal
handler being configured, the failure of the first signal delivery will
result in *another* signal getting delivered. The second signal may
succeed if another thread has resolved the issue that triggered the
segfault (i.e. a well timed mprotect()/mmap()), or the second signal is
being delivered to another stack (i.e. an alt stack).

On x86, in the non-shadow stack case, all the accesses to userspace are
done before changes to the registers (in pt_regs). The operation is
aborted when an access error occurs, so although there may be writes done
for the first signal, control flow changes for the signal (regs->ip,
regs->sp, etc) are not committed until all the accesses have already
completed successfully. This means that the second signal will be
delivered as if it happened at the time of the first signal. It will
effectively replace the first aborted signal, overwriting the half-written
frame of the aborted signal. So on sigreturn from the second signal,
control flow will resume happily from the point of control flow where the
original signal was delivered.

The problem is, when shadow stack is active, the shadow stack SSP
register/MSR is updated *before* some of the userspace accesses. This
means if the earlier accesses succeed and the later ones fail, the second
signal will not be delivered at the same spot on the shadow stack as the
first one. So on sigreturn from the second signal, the SSP will be
pointing to the wrong location on the shadow stack (off by a frame).

Pengfei privately reported that while using a shadow stack enabled glibc,
the “signal06” test in the LTP test-suite hung. It turns out it is
testing the above described double signal scenario. When this test was
compiled with shadow stack, the first signal pushed a shadow stack
sigframe, then the second pushed another. When the second signal was
handled, the SSP was at the first shadow stack signal frame instead of
the original location. The test then got stuck as the #CP from the twice
incremented SSP was incorrect and generated segfaults in a loop.

Fix this by adjusting the SSP register only after any userspace accesses,
such that there can be no failures after the SSP is adjusted. Do this by
moving the shadow stack sigframe push logic to happen after all other
userspace accesses.

Note, sigreturn (as opposed to the signal delivery dealt with in this
patch) has ordering behavior that could lead to similar failures. The
ordering issues there extend beyond shadow stack to include the alt stack
restoration. Fixing that would require cross-arch changes, and the
ordering today does not cause any known test or apps breakages. So leave
it as is, for now.

[ dhansen: minor changelog/subject tweak ]

Fixes: 05e36022c054 ("x86/shstk: Handle signals for shadow stack")
Reported-by: Pengfei Xu <pengfei.xu@intel.com>
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Tested-by: Pengfei Xu <pengfei.xu@intel.com>
Cc:stable@vger.kernel.org
Link: https://lore.kernel.org/all/20231107182251.91276-1-rick.p.edgecombe%40intel.com
Link: https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/syscalls/signal/signal06.c
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: ufs: core: Fix racing issue between ufshcd_mcq_abort() and ISR

commit 27900d7119c464b43cd9eac69c85884d17bae240 upstream.

If command timeout happens and cq complete IRQ is raised at the same time,
ufshcd_mcq_abort clears lprb->cmd and a NULL pointer deref happens in the
ISR. Error log:

ufshcd_abort: Device abort task at tag 18
Unable to handle kernel NULL pointer dereference at virtual address
0000000000000108
pc : [0xffffffe27ef867ac] scsi_dma_unmap+0xc/0x44
lr : [0xffffffe27f1b898c] ufshcd_release_scsi_cmd+0x24/0x114

Fixes: f1304d442077 ("scsi: ufs: mcq: Added ufshcd_mcq_abort()")
Cc: stable@vger.kernel.org
Signed-off-by: Peter Wang <peter.wang@mediatek.com>
Link: https://lore.kernel.org/r/20231106075117.8995-1-peter.wang@mediatek.com
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix system crash due to bad pointer access

commit 19597cad64d608aa8ac2f8aef50a50187a565223 upstream.

User experiences system crash when running AER error injection.  The
perturbation causes the abort-all-I/O path to trigger. The driver assumes
all I/O on this path is FCP only. If there is both NVMe & FCP traffic, a
system crash happens. Add additional check to see if I/O is FCP or not
before access.

PID: 999019  TASK: ff35d769f24722c0  CPU: 53  COMMAND: "kworker/53:1"
 0 [ff3f78b964847b58] machine_kexec at ffffffffae86973d
 1 [ff3f78b964847ba8] __crash_kexec at ffffffffae9be29d
 2 [ff3f78b964847c70] crash_kexec at ffffffffae9bf528
 3 [ff3f78b964847c78] oops_end at ffffffffae8282ab
 4 [ff3f78b964847c98] exc_page_fault at ffffffffaf2da502
 5 [ff3f78b964847cc0] asm_exc_page_fault at ffffffffaf400b62
   [exception RIP: qla2x00_abort_srb+444]
   RIP: ffffffffc07b5f8c  RSP: ff3f78b964847d78  RFLAGS: 00010046
   RAX: 0000000000000282  RBX: ff35d74a0195a200  RCX: ff35d76886fd03a0
   RDX: 0000000000000001  RSI: ffffffffc07c5ec8  RDI: ff35d74a0195a200
   RBP: ff35d76913d22080   R8: ff35d7694d103200   R9: ff35d7694d103200
   R10: 0000000100000000  R11: ffffffffb05d6630  R12: 0000000000010000
   R13: ff3f78b964847df8  R14: ff35d768d8754000  R15: ff35d768877248e0
   ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 6 [ff3f78b964847d70] qla2x00_abort_srb at ffffffffc07b5f84 [qla2xxx]
 7 [ff3f78b964847de0] __qla2x00_abort_all_cmds at ffffffffc07b6238 [qla2xxx]
 8 [ff3f78b964847e38] qla2x00_abort_all_cmds at ffffffffc07ba635 [qla2xxx]
 9 [ff3f78b964847e58] qla2x00_terminate_rport_io at ffffffffc08145eb [qla2xxx]
10 [ff3f78b964847e70] fc_terminate_rport_io at ffffffffc045987e [scsi_transport_fc]
11 [ff3f78b964847e88] process_one_work at ffffffffae914f15
12 [ff3f78b964847ed0] worker_thread at ffffffffae9154c0
13 [ff3f78b964847f10] kthread at ffffffffae91c456
14 [ff3f78b964847f50] ret_from_fork at ffffffffae8036ef

Cc: stable@vger.kernel.org
Fixes: f45bca8c5052 ("scsi: qla2xxx: Fix double scsi_done for abort path")
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Link: https://lore.kernel.org/r/20231030064912.37912-1-njavali@marvell.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: ufs: qcom: Update PHY settings only when scaling to higher gears

commit fc88ca19ad0989dc0e4d4b126d5d0ba91f6cb616 upstream.

The "hs_gear" variable is used to program the PHY settings (submode) during
ufs_qcom_power_up_sequence(). Currently, it is being updated every time the
agreed gear changes. Due to this, if the gear got downscaled before suspend
(runtime/system), then while resuming, the PHY settings for the lower gear
will be applied first and later when scaling to max gear with REINIT, the
PHY settings for the max gear will be applied.

This adds a latency while resuming and also really not needed as the PHY
gear settings are backwards compatible i.e., we can continue using the PHY
settings for max gear with lower gear speed.

So let's update the "hs_gear" variable _only_ when the agreed gear is
greater than the current one. This guarantees that the PHY settings will be
changed only during probe time and fatal error condition.

Due to this, UFSHCD_QUIRK_REINIT_AFTER_MAX_GEAR_SWITCH can now be skipped
when the PM operation is in progress.

Cc: stable@vger.kernel.org
Fixes: 96a7141da332 ("scsi: ufs: core: Add support for reinitializing the UFS device")
Reported-by: Can Guo <quic_cang@quicinc.com>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20230908145329.154024-1-manivannan.sadhasivam@linaro.org
Reviewed-by: Can Guo <quic_cang@quicinc.com>
Tested-by: Can Guo <quic_cang@quicinc.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: megaraid_sas: Increase register read retry rount from 3 to 30 for selected registers

commit 8e3ed9e786511ad800c33605ed904b9de49323cf upstream.

In BMC environments with concurrent access to multiple registers, certain
registers occasionally yield a value of 0 even after 3 retries due to
hardware errata. As a fix, we have extended the retry count from 3 to 30.

The same errata applies to the mpt3sas driver, and a similar patch has
been accepted. Please find more details in the mpt3sas patch reference
link.

Link: https://lore.kernel.org/r/20230829090020.5417-2-ranjan.kumar@broadcom.com
Fixes: 272652fcbf1a ("scsi: megaraid_sas: add retry logic in megasas_readl")
Cc: stable@vger.kernel.org
Signed-off-by: Chandrakanth patil <chandrakanth.patil@broadcom.com>
Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Link: https://lore.kernel.org/r/20231003110021.168862-2-chandrakanth.patil@broadcom.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: mpt3sas: Fix loop logic

commit 3c978492c333f0c08248a8d51cecbe5eb5f617c9 upstream.

The retry loop continues to iterate until the count reaches 30, even after
receiving the correct value. Exit loop when a non-zero value is read.

Fixes: 4ca10f3e3174 ("scsi: mpt3sas: Perform additional retries if doorbell read returns 0")
Cc: stable@vger.kernel.org
Signed-off-by: Ranjan Kumar <ranjan.kumar@broadcom.com>
Link: https://lore.kernel.org/r/20231020105849.6350-1-ranjan.kumar@broadcom.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bpf: Fix precision tracking for BPF_ALU | BPF_TO_BE | BPF_END

commit 291d044fd51f8484066300ee42afecf8c8db7b3a upstream.

BPF_END and BPF_NEG has a different specification for the source bit in
the opcode compared to other ALU/ALU64 instructions, and is either
reserved or use to specify the byte swap endianness. In both cases the
source bit does not encode source operand location, and src_reg is a
reserved field.

backtrack_insn() currently does not differentiate BPF_END and BPF_NEG
from other ALU/ALU64 instructions, which leads to r0 being incorrectly
marked as precise when processing BPF_ALU | BPF_TO_BE | BPF_END
instructions. This commit teaches backtrack_insn() to correctly mark
precision for such case.

While precise tracking of BPF_NEG and other BPF_END instructions are
correct and does not need fixing, this commit opt to process all BPF_NEG
and BPF_END instructions within the same if-clause to better align with
current convention used in the verifier (e.g. check_alu_op).

Fixes: b5dc0163d8fd ("bpf: precise scalar_value tracking")
Cc: stable@vger.kernel.org
Reported-by: Mohamed Mahmoud <mmahmoud@redhat.com>
Closes: https://lore.kernel.org/r/87jzrrwptf.fsf@toke.dk
Tested-by: Toke Høiland-Jørgensen <toke@redhat.com>
Tested-by: Tao Lyu <tao.lyu@epfl.ch>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Link: https://lore.kernel.org/r/20231102053913.12004-2-shung-hsi.yu@suse.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bpf: Fix check_stack_write_fixed_off() to correctly spill imm

commit 811c363645b33e6e22658634329e95f383dfc705 upstream.

In check_stack_write_fixed_off(), imm value is cast to u32 before being
spilled to the stack. Therefore, the sign information is lost, and the
range information is incorrect when load from the stack again.

For the following prog:
0: r2 = r10
1: *(u64*)(r2 -40) = -44
2: r0 = *(u64*)(r2 - 40)
3: if r0 s<= 0xa goto +2
4: r0 = 1
5: exit
6: r0  = 0
7: exit

The verifier gives:
func#0 @0
0: R1=ctx(off=0,imm=0) R10=fp0
0: (bf) r2 = r10                      ; R2_w=fp0 R10=fp0
1: (7a) *(u64 *)(r2 -40) = -44        ; R2_w=fp0 fp-40_w=4294967252
2: (79) r0 = *(u64 *)(r2 -40)         ; R0_w=4294967252 R2_w=fp0
fp-40_w=4294967252
3: (c5) if r0 s< 0xa goto pc+2
mark_precise: frame0: last_idx 3 first_idx 0 subseq_idx -1
mark_precise: frame0: regs=r0 stack= before 2: (79) r0 = *(u64 *)(r2 -40)
3: R0_w=4294967252
4: (b7) r0 = 1                        ; R0_w=1
5: (95) exit
verification time 7971 usec
stack depth 40
processed 6 insns (limit 1000000) max_states_per_insn 0 total_states 0
peak_states 0 mark_read 0

So remove the incorrect cast, since imm field is declared as s32, and
__mark_reg_known() takes u64, so imm would be correctly sign extended
by compiler.

Fixes: ecdf985d7615 ("bpf: track immediate values written to stack by BPF_ST instruction")
Cc: stable@vger.kernel.org
Signed-off-by: Hao Sun <sunhao.th@gmail.com>
Acked-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20231101-fix-check-stack-write-v3-1-f05c2b1473d5@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

spi: Fix null dereference on suspend

commit bef4a48f4ef798c4feddf045d49e53c8a97d5e37 upstream.

A race condition exists where a synchronous (noqueue) transfer can be
active during a system suspend. This can cause a null pointer
dereference exception to occur when the system resumes.

Example order of events leading to the exception:
1. spi_sync() calls __spi_transfer_message_noqueue() which sets
   ctlr->cur_msg
2. Spi transfer begins via spi_transfer_one_message()
3. System is suspended interrupting the transfer context
4. System is resumed
6. spi_controller_resume() calls spi_start_queue() which resets cur_msg
   to NULL
7. Spi transfer context resumes and spi_finalize_current_message() is
   called which dereferences cur_msg (which is now NULL)

Wait for synchronous transfers to complete before suspending by
acquiring the bus mutex and setting/checking a suspend flag.

Signed-off-by: Mark Hasemeyer <markhas@chromium.org>
Link: https://lore.kernel.org/r/20231107144743.v1.1.I7987f05f61901f567f7661763646cb7d7919b528@changeid
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

randstruct: Fix gcc-plugin performance mode to stay in group

commit 381fdb73d1e2a48244de7260550e453d1003bb8e upstream.

The performance mode of the gcc-plugin randstruct was shuffling struct
members outside of the cache-line groups. Limit the range to the
specified group indexes.

Cc: linux-hardening@vger.kernel.org
Cc: stable@vger.kernel.org
Reported-by: Lukas Loidolt <e1634039@student.tuwien.ac.at>
Closes: https://lore.kernel.org/all/f3ca77f0-e414-4065-83a5-ae4c4d25545d@student.tuwien.ac.at
Fixes: 313dd1b62921 ("gcc-plugins: Add the randstruct plugin")
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powerpc/perf: Fix disabling BHRB and instruction sampling

commit ea142e590aec55ba40c5affb4d49e68c713c63dc upstream.

When the PMU is disabled, MMCRA is not updated to disable BHRB and
instruction sampling. This can lead to those features remaining enabled,
which can slow down a real or emulated CPU.

Fixes: 1cade527f6e9 ("powerpc/perf: BHRB control to disable BHRB logic when not used")
Cc: stable@vger.kernel.org # v5.9+
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20231018153423.298373-1-npiggin@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf intel-pt: Fix async branch flags

commit f2d87895cbc4af80649850dcf5da36de6b2ed3dd upstream.

Ensure PERF_IP_FLAG_ASYNC is set always for asynchronous branches (i.e.
interrupts etc).

Fixes: 90e457f7be08 ("perf tools: Add Intel PT support")
Cc: stable@vger.kernel.org
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20230928072953.19369-1-adrian.hunter@intel.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: venus: hfi: add checks to perform sanity on queue pointers

commit 5e538fce33589da6d7cb2de1445b84d3a8a692f7 upstream.

Read and write pointers are used to track the packet index in the memory
shared between video driver and firmware. There is a possibility of OOB
access if the read or write pointer goes beyond the queue memory size.
Add checks for the read and write pointer to avoid OOB access.

Cc: stable@vger.kernel.org
Fixes: d96d3f30c0f2 ("[media] media: venus: hfi: add Venus HFI files")
Signed-off-by: Vikash Garodia <quic_vgarodia@quicinc.com>
Signed-off-by: Stanimir Varbanov <stanimir.k.varbanov@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers: perf: Check find_first_bit() return value

commit c6e316ac05532febb0c966fa9b55f5258ed037be upstream.

We must check the return value of find_first_bit() before using the
return value as an index array since it happens to overflow the array
and then panic:

[  107.318430] Kernel BUG [#1]
[  107.319434] CPU: 3 PID: 1238 Comm: kill Tainted: G            E      6.6.0-rc6ubuntu-defconfig #2
[  107.319465] Hardware name: riscv-virtio,qemu (DT)
[  107.319551] epc : pmu_sbi_ovf_handler+0x3a4/0x3ae
[  107.319840]  ra : pmu_sbi_ovf_handler+0x52/0x3ae
[  107.319868] epc : ffffffff80a0a77c ra : ffffffff80a0a42a sp : ffffaf83fecda350
[  107.319884]  gp : ffffffff823961a8 tp : ffffaf8083db1dc0 t0 : ffffaf83fecda480
[  107.319899]  t1 : ffffffff80cafe62 t2 : 000000000000ff00 s0 : ffffaf83fecda520
[  107.319921]  s1 : ffffaf83fecda380 a0 : 00000018fca29df0 a1 : ffffffffffffffff
[  107.319936]  a2 : 0000000001073734 a3 : 0000000000000004 a4 : 0000000000000000
[  107.319951]  a5 : 0000000000000040 a6 : 000000001d1c8774 a7 : 0000000000504d55
[  107.319965]  s2 : ffffffff82451f10 s3 : ffffffff82724e70 s4 : 000000000000003f
[  107.319980]  s5 : 0000000000000011 s6 : ffffaf8083db27c0 s7 : 0000000000000000
[  107.319995]  s8 : 0000000000000001 s9 : 00007fffb45d6558 s10: 00007fffb45d81a0
[  107.320009]  s11: ffffaf7ffff60000 t3 : 0000000000000004 t4 : 0000000000000000
[  107.320023]  t5 : ffffaf7f80000000 t6 : ffffaf8000000000
[  107.320037] status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000003
[  107.320081] [<ffffffff80a0a77c>] pmu_sbi_ovf_handler+0x3a4/0x3ae
[  107.320112] [<ffffffff800b42d0>] handle_percpu_devid_irq+0x9e/0x1a0
[  107.320131] [<ffffffff800ad92c>] generic_handle_domain_irq+0x28/0x36
[  107.320148] [<ffffffff8065f9f8>] riscv_intc_irq+0x36/0x4e
[  107.320166] [<ffffffff80caf4a0>] handle_riscv_irq+0x54/0x86
[  107.320189] [<ffffffff80cb0036>] do_irq+0x64/0x96
[  107.320271] Code: 85a6 855e b097 ff7f 80e7 9220 b709 9002 4501 bbd9 (9002) 6097
[  107.320585] ---[ end trace 0000000000000000 ]---
[  107.320704] Kernel panic - not syncing: Fatal exception in interrupt
[  107.320775] SMP: stopping secondary CPUs
[  107.321219] Kernel Offset: 0x0 from 0xffffffff80000000
[  107.333051] ---[ end Kernel panic - not syncing: Fatal exception in interrupt ]---

Fixes: 4905ec2fb7e6 ("RISC-V: Add sscofpmf extension support")
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20231109082128.40777-1-alexghiti@rivosinc.com
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf: arm_cspmu: Reject events meant for other PMUs

commit 15c7ef7341a2e54cfa12ac502c65d6fd2cce2b62 upstream.

Coresight PMU driver didn't reject events meant for other PMUs.
This caused some of the Core PMU events disappearing from
the output of "perf list". In addition, trying to run e.g.

     $ perf stat -e r2 sleep 1

made Coresight PMU driver to handle the event instead of letting
Core PMU driver to deal with it.

Cc: stable@vger.kernel.org
Fixes: e37dfd65731d ("perf: arm_cspmu: Add support for ARM CoreSight PMU driver")
Signed-off-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
Acked-by: Will Deacon <will@kernel.org>
Reviewed-by: Besar Wicaksono <bwicaksono@nvidia.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20231103001654.35565-1-ilkka@os.amperecomputing.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

i915/perf: Fix NULL deref bugs with drm_dbg() calls

commit 471aa951bf1206d3c10d0daa67005b8e4db4ff83 upstream.

When i915 perf interface is not available dereferencing it will lead to
NULL dereferences.

As returning -ENOTSUPP is pretty clear return when perf interface is not
available.

Fixes: 2fec539112e8 ("i915/perf: Replace DRM_DEBUG with driver specific drm_dbg call")
Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: <stable@vger.kernel.org> # v6.0+
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231027172822.2753059-1-harshit.m.mogalapalli@oracle.com
[tursulin: added stable tag]
(cherry picked from commit 36f27350ff745bd228ab04d7845dfbffc177a889)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

perf/core: Fix cpuctx refcounting

commit 889c58b3155ff4c8e8671c95daef63d6fabbb6b1 upstream.

Audit of the refcounting turned up that perf_pmu_migrate_context()
fails to migrate the ctx refcount.

Fixes: bd2756811766 ("perf: Rewrite core context handling")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lkml.kernel.org/r/20230612093539.085862001@infradead.org
Cc: <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cifs: fix check of rc in function generate_smb3signingkey

[ Upstream commit 181724fc72486dec2bec8803459be05b5162aaa8 ]

Remove extra check after condition, add check after generating key
for encryption. The check is needed to return non zero rc before
rewriting it with generating key for decryption.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com>
Fixes: d70e9fa55884 ("cifs: try opening channels after mounting")
Signed-off-by: Ekaterina Esina <eesina@astralinux.ru>
Co-developed-by: Anastasia Belova <abelova@astralinux.ru>
Signed-off-by: Anastasia Belova <abelova@astralinux.ru>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cifs: spnego: add ';' in HOST_KEY_LEN

[ Upstream commit ff31ba19d732efb9aca3633935d71085e68d5076 ]

"host=" should start with ';' (as in cifs_get_spnego_key)
So its length should be 6.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com>
Fixes: 7c9c3760b3a5 ("[CIFS] add constants for string lengths of keynames in SPNEGO upcall string")
Signed-off-by: Anastasia Belova <abelova@astralinux.ru>
Co-developed-by: Ekaterina Esina <eesina@astralinux.ru>
Signed-off-by: Ekaterina Esina <eesina@astralinux.ru>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: ufs: core: Expand MCQ queue slot to DeviceQueueDepth + 1

[ Upstream commit defde5a50d91c74e1ce71a7f0bce7fb1ae311d84 ]

The UFSHCI 4.0 specification mandates that there should always be at least
one empty slot in each queue for distinguishing between full and empty
states. Enlarge 'hwq->max_entries' to 'DeviceQueueDepth + 1' to allow
UFSHCI 4.0 controllers to fully utilize MCQ queue slots.

Fixes: 4682abfae2eb ("scsi: ufs: core: mcq: Allocate memory for MCQ mode")
Signed-off-by: Naomi Chu <naomi.chu@mediatek.com>
Link: https://lore.kernel.org/r/20231102052426.12006-2-naomi.chu@mediatek.com
Reviewed-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Peter Wang <peter.wang@mediatek.com>
Reviewed-by: Chun-Hung <chun-hung.wu@mediatek.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tools/power/turbostat: Enable the C-state Pre-wake printing

[ Upstream commit b61b7d8c4c22c4298a50ae5d0ee88facb85ce665 ]

Currently the C-state Pre-wake will not be printed due to the
probe has not been invoked. Invoke the probe function accordingly.

Fixes: aeb01e6d71ff ("tools/power turbostat: Print the C-state Pre-wake settings")
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Reviewed-by: Len Brown <len.brown@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tools/power/turbostat: Fix a knl bug

[ Upstream commit 137f01b3529d292a68d22e9681e2f903c768f790 ]

MSR_KNL_CORE_C6_RESIDENCY should be evaluated only if
1. this is KNL platform
AND
2. need to get C6 residency or need to calculate C1 residency

Fix the broken logic introduced by commit 1e9042b9c8d4 ("tools/power
turbostat: Fix CPU%C1 display value").

Fixes: 1e9042b9c8d4 ("tools/power turbostat: Fix CPU%C1 display value")
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Reviewed-by: Len Brown <len.brown@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

macvlan: Don't propagate promisc change to lower dev in passthru

[ Upstream commit 7e1caeace0418381f36b3aa8403dfd82fc57fc53 ]

Macvlan device in passthru mode sets its lower device promiscuous mode
according to its MACVLAN_FLAG_NOPROMISC flag instead of synchronizing it to
its own promiscuity setting. However, macvlan_change_rx_flags() function
doesn't check the mode before propagating such changes to the lower device
which can cause net_device->promiscuity counter overflow as illustrated by
reproduction example [0] and resulting dmesg log [1]. Fix the issue by
first verifying the mode in macvlan_change_rx_flags() function before
propagating promiscuous mode change to the lower device.

[0]:
ip link add macvlan1 link enp8s0f0 type macvlan mode passthru
ip link set macvlan1 promisc on
ip l set dev macvlan1 up
ip link set macvlan1 promisc off
ip l set dev macvlan1 down
ip l set dev macvlan1 up

[1]:
[ 5156.281724] macvlan1: entered promiscuous mode
[ 5156.285467] mlx5_core 0000:08:00.0 enp8s0f0: entered promiscuous mode
[ 5156.287639] macvlan1: left promiscuous mode
[ 5156.288339] mlx5_core 0000:08:00.0 enp8s0f0: left promiscuous mode
[ 5156.290907] mlx5_core 0000:08:00.0 enp8s0f0: entered promiscuous mode
[ 5156.317197] mlx5_core 0000:08:00.0 enp8s0f0: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.

Fixes: efdbd2b30caa ("macvlan: Propagate promiscuity setting to lower devices.")
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20231114175915.1649154-1-vladbu@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: sched: do not offload flows with a helper in act_ct

[ Upstream commit 7cd5af0e937a197295f3aa3721031f0fbae49cff ]

There is no hardware supporting ct helper offload. However, prior to this
patch, a flower filter with a helper in the ct action can be successfully
set into the HW, for example (eth1 is a bnxt NIC):

  # tc qdisc add dev eth1 ingress_block 22 ingress
  # tc filter add block 22 proto ip flower skip_sw ip_proto tcp \
    dst_port 21 ct_state -trk action ct helper ipv4-tcp-ftp
  # tc filter show dev eth1 ingress

    filter block 22 protocol ip pref 49152 flower chain 0 handle 0x1
      eth_type ipv4
      ip_proto tcp
      dst_port 21
      ct_state -trk
      skip_sw
      in_hw in_hw_count 1   <----
        action order 1: ct zone 0 helper ipv4-tcp-ftp pipe
         index 2 ref 1 bind 1
        used_hw_stats delayed

This might cause the flower filter not to work as expected in the HW.

This patch avoids this problem by simply returning -EOPNOTSUPP in
tcf_ct_offload_act_setup() to not allow to offload flows with a helper
in act_ct.

Fixes: a21b06e73191 ("net: sched: add helper support in act_ct")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/f8685ec7702c4a448a1371a8b34b43217b583b9d.1699898008.git.lucien.xin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Check return value of snprintf writing to fw_version buffer for representors

[ Upstream commit 1b2bd0c0264febcd8d47209079a6671c38e6558b ]

Treat the operation as an error case when the return value is equivalent to
the size of the name buffer. Failed to write null terminator to the name
buffer, making the string malformed and should not be used. Provide a
string with only the firmware version when forming the string with the
board id fails. This logic for representors is identical to normal flow
with ethtool.

Without check, will trigger -Wformat-truncation with W=1.

    drivers/net/ethernet/mellanox/mlx5/core/en_rep.c: In function 'mlx5e_rep_get_drvinfo':
    drivers/net/ethernet/mellanox/mlx5/core/en_rep.c:78:31: warning: '%.16s' directive output may be truncated writing up to 16 bytes into a region of size between 13 and 22 [-Wformat-truncation=]
      78 |                  "%d.%d.%04d (%.16s)",
         |                               ^~~~~
    drivers/net/ethernet/mellanox/mlx5/core/en_rep.c:77:9: note: 'snprintf' output between 12 and 37 bytes into a destination of size 32
      77 |         snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version),
         |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      78 |                  "%d.%d.%04d (%.16s)",
         |                  ~~~~~~~~~~~~~~~~~~~~~
      79 |                  fw_rev_maj(mdev), fw_rev_min(mdev),
         |                  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      80 |                  fw_rev_sub(mdev), mdev->board_id);
         |                  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Fixes: cf83c8fdcd47 ("net/mlx5e: Add missing ethtool driver info for representors")
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=6d4ab2e97dcfbcd748ae71761a9d8e5e41cc732c
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-16-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Check return value of snprintf writing to fw_version buffer

[ Upstream commit 41e63c2baa11dc2aa71df5dd27a5bd87d11b6bbb ]

Treat the operation as an error case when the return value is equivalent to
the size of the name buffer. Failed to write null terminator to the name
buffer, making the string malformed and should not be used. Provide a
string with only the firmware version when forming the string with the
board id fails.

Without check, will trigger -Wformat-truncation with W=1.

    drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c: In function 'mlx5e_ethtool_get_drvinfo':
    drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c:49:31: warning: '%.16s' directive output may be truncated writing up to 16 bytes into a region of size between 13 and 22 [-Wformat-truncation=]
      49 |                  "%d.%d.%04d (%.16s)",
         |                               ^~~~~
    drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c:48:9: note: 'snprintf' output between 12 and 37 bytes into a destination of size 32
      48 |         snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version),
         |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      49 |                  "%d.%d.%04d (%.16s)",
         |                  ~~~~~~~~~~~~~~~~~~~~~
      50 |                  fw_rev_maj(mdev), fw_rev_min(mdev), fw_rev_sub(mdev),
         |                  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      51 |                  mdev->board_id);
         |                  ~~~~~~~~~~~~~~~

Fixes: 84e11edb71de ("net/mlx5e: Show board id in ethtool driver information")
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=6d4ab2e97dcfbcd748ae71761a9d8e5e41cc732c
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Reduce the size of icosq_str

[ Upstream commit dce94142842e119b982c27c1b62bd20890c7fd21 ]

icosq_str size is unnecessarily too long, and it causes a build warning
-Wformat-truncation with W=1. Looking closely, It doesn't need to be 255B,
hence this patch reduces the size to 32B which should be more than enough
to host the string: "ICOSQ: 0x%x, ".

While here, add a missing space in the formatted string.

This fixes the following build warning:

$ KCFLAGS='-Wall -Werror'
$ make O=/tmp/kbuild/linux W=1 -s -j12 drivers/net/ethernet/mellanox/mlx5/core/

drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c: In function 'mlx5e_reporter_rx_timeout':
drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c:718:56:
error: ', CQ: 0x' directive output may be truncated writing 8 bytes into a region of size between 0 and 255 [-Werror=format-truncation=]
  718 |                  "RX timeout on channel: %d, %sRQ: 0x%x, CQ: 0x%x",
      |                                                        ^~~~~~~~
drivers/net/ethernet/mellanox/mlx5/core/en/reporter_rx.c:717:9: note: 'snprintf' output between 43 and 322 bytes into a destination of size 288
  717 |         snprintf(err_str, sizeof(err_str),
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  718 |                  "RX timeout on channel: %d, %sRQ: 0x%x, CQ: 0x%x",
      |                  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  719 |                  rq->ix, icosq_str, rq->rqn, rq->cq.mcq.cqn);
      |                  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Fixes: 521f31af004a ("net/mlx5e: Allow RQ outside of channel context")
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=6d4ab2e97dcfbcd748ae71761a9d8e5e41cc732c
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-14-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Increase size of irq name buffer

[ Upstream commit 3338bebfc26a1e2cebbba82a1cf12c0159608e73 ]

Without increased buffer size, will trigger -Wformat-truncation with W=1
for the snprintf operation writing to the buffer.

    drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c: In function 'mlx5_irq_alloc':
    drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c:296:7: error: '@pci:' directive output may be truncated writing 5 bytes into a region of size between 1 and 32 [-Werror=format-truncation=]
      296 |    "%s@pci:%s", name, pci_name(dev->pdev));
          |       ^~~~~
    drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c:295:2: note: 'snprintf' output 6 or more bytes (assuming 37) into a destination of size 32
      295 |  snprintf(irq->name, MLX5_MAX_IRQ_NAME,
          |  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      296 |    "%s@pci:%s", name, pci_name(dev->pdev));
          |    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Fixes: ada9f5d00797 ("IB/mlx5: Fix eq names to display nicely in /proc/interrupts")
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=6d4ab2e97dcfbcd748ae71761a9d8e5e41cc732c
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-13-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Update doorbell for port timestamping CQ before the software counter

[ Upstream commit 92214be5979c0961a471b7eaaaeacab41bdf456c ]

Previously, mlx5e_ptp_poll_ts_cq would update the device doorbell with the
incremented consumer index after the relevant software counters in the
kernel were updated. In the mlx5e_sq_xmit_wqe context, this would lead to
either overrunning the device CQ or exceeding the expected software buffer
size in the device CQ if the device CQ size was greater than the software
buffer size. Update the relevant software counter only after updating the
device CQ consumer index in the port timestamping napi_poll context.

Log:
    mlx5_core 0000:08:00.0: cq_err_event_notifier:517:(pid 0): CQ error on CQN 0x487, syndrome 0x1
    mlx5_core 0000:08:00.0 eth2: mlx5e_cq_error_event: cqn=0x000487 event=0x04

Fixes: 1880bc4e4a96 ("net/mlx5e: Add TX port timestamp support")
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-12-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Track xmit submission to PTP WQ after populating metadata map

[ Upstream commit 7e3f3ba97e6cc6fce5bf62df2ca06c8e59040167 ]

Ensure the skb is available in metadata mapping to skbs before tracking the
metadata index for detecting undelivered CQEs. If the metadata index is put
in the tracking list before putting the skb in the map, the metadata index
might be used for detecting undelivered CQEs before the relevant skb is
available in the map, which can lead to a null-ptr-deref.

Log:
    general protection fault, probably for non-canonical address 0xdffffc0000000005: 0000 [#1] SMP KASAN
    KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]
    CPU: 0 PID: 1243 Comm: kworker/0:2 Not tainted 6.6.0-rc4+ #108
    Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
    Workqueue: events mlx5e_rx_dim_work [mlx5_core]
    RIP: 0010:mlx5e_ptp_napi_poll+0x9a4/0x2290 [mlx5_core]
    Code: 8c 24 38 cc ff ff 4c 8d 3c c1 4c 89 f9 48 c1 e9 03 42 80 3c 31 00 0f 85 97 0f 00 00 4d 8b 3f 49 8d 7f 28 48 89 f9 48 c1 e9 03 <42> 80 3c 31 00 0f 85 8b 0f 00 00 49 8b 47 28 48 85 c0 0f 84 05 07
    RSP: 0018:ffff8884d3c09c88 EFLAGS: 00010206
    RAX: 0000000000000069 RBX: ffff8881160349d8 RCX: 0000000000000005
    RDX: ffffed10218f48cf RSI: 0000000000000004 RDI: 0000000000000028
    RBP: ffff888122707700 R08: 0000000000000001 R09: ffffed109a781383
    R10: 0000000000000003 R11: 0000000000000003 R12: ffff88810c7a7a40
    R13: ffff888122707700 R14: dffffc0000000000 R15: 0000000000000000
    FS:  0000000000000000(0000) GS:ffff8884d3c00000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    CR2: 00007f4f878dd6e0 CR3: 000000014d108002 CR4: 0000000000370eb0
    DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
    DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
    Call Trace:
    <IRQ>
    ? die_addr+0x3c/0xa0
    ? exc_general_protection+0x144/0x210
    ? asm_exc_general_protection+0x22/0x30
    ? mlx5e_ptp_napi_poll+0x9a4/0x2290 [mlx5_core]
    ? mlx5e_ptp_napi_poll+0x8f6/0x2290 [mlx5_core]
    __napi_poll.constprop.0+0xa4/0x580
    net_rx_action+0x460/0xb80
    ? _raw_spin_unlock_irqrestore+0x32/0x60
    ? __napi_poll.constprop.0+0x580/0x580
    ? tasklet_action_common.isra.0+0x2ef/0x760
    __do_softirq+0x26c/0x827
    irq_exit_rcu+0xc2/0x100
    common_interrupt+0x7f/0xa0
    </IRQ>
    <TASK>
    asm_common_interrupt+0x22/0x40
    RIP: 0010:__kmem_cache_alloc_node+0xb/0x330
    Code: 41 5d 41 5e 41 5f c3 8b 44 24 14 8b 4c 24 10 09 c8 eb d5 e8 b7 43 ca 01 0f 1f 80 00 00 00 00 0f 1f 44 00 00 55 48 89 e5 41 57 <41> 56 41 89 d6 41 55 41 89 f5 41 54 49 89 fc 53 48 83 e4 f0 48 83
    RSP: 0018:ffff88812c4079c0 EFLAGS: 00000246
    RAX: 1ffffffff083c7fe RBX: ffff888100042dc0 RCX: 0000000000000218
    RDX: 00000000ffffffff RSI: 0000000000000dc0 RDI: ffff888100042dc0
    RBP: ffff88812c4079c8 R08: ffffffffa0289f96 R09: ffffed1025880ea9
    R10: ffff888138839f80 R11: 0000000000000002 R12: 0000000000000dc0
    R13: 0000000000000100 R14: 000000000000008c R15: ffff8881271fc450
    ? cmd_exec+0x796/0x2200 [mlx5_core]
    kmalloc_trace+0x26/0xc0
    cmd_exec+0x796/0x2200 [mlx5_core]
    mlx5_cmd_do+0x22/0xc0 [mlx5_core]
    mlx5_cmd_exec+0x17/0x30 [mlx5_core]
    mlx5_core_modify_cq_moderation+0x139/0x1b0 [mlx5_core]
    ? mlx5_add_cq_to_tasklet+0x280/0x280 [mlx5_core]
    ? lockdep_set_lock_cmp_fn+0x190/0x190
    ? process_one_work+0x659/0x1220
    mlx5e_rx_dim_work+0x9d/0x100 [mlx5_core]
    process_one_work+0x730/0x1220
    ? lockdep_hardirqs_on_prepare+0x400/0x400
    ? max_active_store+0xf0/0xf0
    ? assign_work+0x168/0x240
    worker_thread+0x70f/0x12d0
    ? __kthread_parkme+0xd1/0x1d0
    ? process_one_work+0x1220/0x1220
    kthread+0x2d9/0x3b0
    ? kthread_complete_and_exit+0x20/0x20
    ret_from_fork+0x2d/0x70
    ? kthread_complete_and_exit+0x20/0x20
    ret_from_fork_asm+0x11/0x20
    </TASK>
    Modules linked in: xt_conntrack xt_MASQUERADE nf_conntrack_netlink nfnetlink xt_addrtype iptable_nat nf_nat br_netfilter rpcsec_gss_krb5 auth_rpcgss oid_registry overlay mlx5_ib ib_uverbs ib_core zram zsmalloc mlx5_core fuse
    ---[ end trace 0000000000000000 ]---

Fixes: 3178308ad4ca ("net/mlx5e: Make tx_port_ts logic resilient to out-of-order CQEs")
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-11-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Avoid referencing skb after free-ing in drop path of mlx5e_sq_xmit_wqe

[ Upstream commit 64f14d16eef1f939000f2617b50c7c996b5117d4 ]

When SQ is a port timestamping SQ for PTP, do not access tx flags of skb
after free-ing the skb. Free the skb only after all references that depend
on it have been handled in the dropped WQE path.

Fixes: 3178308ad4ca ("net/mlx5e: Make tx_port_ts logic resilient to out-of-order CQEs")
Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-10-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Don't modify the peer sent-to-vport rules for IPSec offload

[ Upstream commit bdf788cf224f61c20a01c58c00685d394d57887f ]

As IPSec packet offload in switchdev mode is not supported with LAG,
it's unnecessary to modify those sent-to-vport rules to the peer eswitch.

Fixes: c6c2bf5db4ea ("net/mlx5e: Support IPsec packet offload for TX in switchdev mode")
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-9-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Fix pedit endianness

[ Upstream commit 0c101a23ca7eaf00eef1328eefb04b3a93401cc8 ]

Referenced commit addressed endianness issue in mlx5 pedit implementation
in ad hoc manner instead of systematically treating integer values
according to their types which left pedit fields of sizes not equal to 4
and where the bytes being modified are not least significant ones broken on
big endian machines since wrong bits will be consumed during parsing which
leads to following example error when applying pedit to source and
destination MAC addresses:

[Wed Oct 18 12:52:42 2023] mlx5_core 0001:00:00.1 p1v3_r: attempt to offload an unsupported field (cmd 0)
[Wed Oct 18 12:52:42 2023] mask: 00000000330c5b68: 00 00 00 00 ff ff 00 00 00 00 ff ff 00 00 00 00  ................
[Wed Oct 18 12:52:42 2023] mask: 0000000017d22fd9: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[Wed Oct 18 12:52:42 2023] mask: 000000008186d717: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[Wed Oct 18 12:52:42 2023] mask: 0000000029eb6149: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[Wed Oct 18 12:52:42 2023] mask: 000000007ed103e4: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[Wed Oct 18 12:52:42 2023] mask: 00000000db8101a6: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
[Wed Oct 18 12:52:42 2023] mask: 00000000ec3c08a9: 00 00 00 00 00 00 00 00 00 00 00 00              ............

Treat masks and values of pedit and filter match as network byte order,
refactor pointers to them to void pointers instead of confusing u32
pointers and only cast to pointer-to-integer when reading a value from
them. Treat pedit mlx5_fields->field_mask as host byte order according to
its type u32, change the constants in fields array accordingly.

Fixes: 82198d8bcdef ("net/mlx5e: Fix endianness when calculating pedit mask first bit")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-8-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: fix double free of encap_header in update funcs

[ Upstream commit 3a4aa3cb83563df942be49d145ee3b7ddf17d6bb ]

Follow up to the previous patch to fix the same issue for
mlx5e_tc_tun_update_header_ipv4{6} when mlx5_packet_reformat_alloc()
fails.

When mlx5_packet_reformat_alloc() fails, the encap_header allocated in
mlx5e_tc_tun_update_header_ipv4{6} will be released within it. However,
e->encap_header is already set to the previously freed encap_header
before mlx5_packet_reformat_alloc(). As a result, the later
mlx5e_encap_put() will free e->encap_header again, causing a double free
issue.

mlx5e_encap_put()
     --> mlx5e_encap_dealloc()
         --> kfree(e->encap_header)

This patch fix it by not setting e->encap_header until
mlx5_packet_reformat_alloc() success.

Fixes: a54e20b4fcae ("net/mlx5e: Add basic TC tunnel set action for SRIOV offloads")
Signed-off-by: Gavin Li <gavinl@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20231114215846.5902-7-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: fix double free of encap_header

[ Upstream commit 6f9b1a0731662648949a1c0587f6acb3b7f8acf1 ]

When mlx5_packet_reformat_alloc() fails, the encap_header allocated in
mlx5e_tc_tun_create_header_ipv4{6} will be released within it. However,
e->encap_header is already set to the previously freed encap_header
before mlx5_packet_reformat_alloc(). As a result, the later
mlx5e_encap_put() will free e->encap_header again, causing a double free
issue.

mlx5e_encap_put()
    --> mlx5e_encap_dealloc()
        --> kfree(e->encap_header)

This happens when cmd: MLX5_CMD_OP_ALLOC_PACKET_REFORMAT_CONTEXT fail.

This patch fix it by not setting e->encap_header until
mlx5_packet_reformat_alloc() success.

Fixes: d589e785baf5e ("net/mlx5e: Allow concurrent creation of encap entries")
Reported-by: Cruz Zhao <cruzzhao@linux.alibaba.com>
Reported-by: Tianchen Ding <dtcccc@linux.alibaba.com>
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>