| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <85653b8a-d0f0-1f3a-ade4-6fe5e42d3d9a@arm.com>
Date: Thu, 22 Jun 2017 20:20:15 +0100
From: Marc Zyngier <marc.zyngier@....com>
To: Shanker Donthineni <shankerd@...eaurora.org>,
linux-kernel <linux-kernel@...r.kernel.org>,
linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>
Cc: Thomas Gleixner <tglx@...utronix.de>,
Jason Cooper <jason@...edaemon.net>,
Vikram Sethi <vikrams@...eaurora.org>
Subject: Re: [PATCH v2] irqchip: gicv3-its: Don't assume GICv3 hardware
supports 16bit INTID
On 22/06/17 18:44, Shanker Donthineni wrote:
> The current ITS driver is assuming every ITS hardware implementation
> supports minimum of 16bit INTID. But this is not true, as per GICv3
> specification, INTID field is IMPLEMENTATION DEFINED in the range of
> 14-24 bits. We might see an unpredictable system behavior on systems
> where hardware support less than 16bits and software tries to use
> 64K LPI interrupts.
>
> On Qualcomm Datacenter Technologies QDF2400 platform, boot log shows
> confusing information about number of LPI chunks as shown below. The
> QDF2400 ITS hardware supports 24bit INTID.
>
> This patch allocates the memory resources for PEND/PROP tables based
> on discoverable value which is specified in GITS_TYPER.IDbits. Also
> taking this opportunity to increase number of LPI/MSI(x) to 128K if
> the hardware is capable, and show log message that reflects the
> correct number of LPI chunks.
>
> ITS@...f7efe0000: allocated 524288 Devices @3c0400000 (indirect, esz 8, psz 64K, shr 1)
> ITS@...f7efe0000: allocated 8192 Interrupt Collections @3c0130000 (flat, esz 8, psz 64K, shr 1)
> ITS@...f7efe0000: allocated 8192 Virtual CPUs @3c0140000 (flat, esz 8, psz 64K, shr 1)
> ITS: Allocated 524032 chunks for LPIs
> PCI/MSI: ITS@...f7efe0000 domain created
> Platform MSI: ITS@...f7efe0000 domain created
I seriously doubt that anyone will ever see a shortage of LPIs with
16bit IDs ("64k interrupts should be enough for everybody"). But if you
think otherwise, fair enough. Comments on the actual patch:
>
> Signed-off-by: Shanker Donthineni <shankerd@...eaurora.org>
> ---
> Changes since v1:
> No code changes, just rebase on tip of the Marc's branch and tested on QDF2400 platform.
> https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/log/?h=irq/irqchip-4.13
>
> drivers/irqchip/irq-gic-v3-its.c | 34 ++++++++++++++++------------------
> 1 file changed, 16 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
> index fee7d13..6000c56 100644
> --- a/drivers/irqchip/irq-gic-v3-its.c
> +++ b/drivers/irqchip/irq-gic-v3-its.c
> @@ -691,9 +691,11 @@ static void its_irq_compose_msi_msg(struct irq_data *d, struct msi_msg *msg)
> */
> #define IRQS_PER_CHUNK_SHIFT 5
> #define IRQS_PER_CHUNK (1 << IRQS_PER_CHUNK_SHIFT)
> +#define ITS_MAX_LPI_NRBITS (17) /* 128K LPIs */
>
> static unsigned long *lpi_bitmap;
> static u32 lpi_chunks;
> +static u32 lpi_nrbits;
> static DEFINE_SPINLOCK(lpi_lock);
>
> static int its_lpi_to_chunk(int lpi)
> @@ -789,26 +791,19 @@ static void its_lpi_free(struct event_lpi_map *map)
> }
>
> /*
> - * We allocate 64kB for PROPBASE. That gives us at most 64K LPIs to
> + * We allocate memory for PROPBASE to cover 2 ^ lpi_nrbits LPIs to
> * deal with (one configuration byte per interrupt). PENDBASE has to
> * be 64kB aligned (one bit per LPI, plus 8192 bits for SPI/PPI/SGI).
> */
> -#define LPI_PROPBASE_SZ SZ_64K
> -#define LPI_PENDBASE_SZ (LPI_PROPBASE_SZ / 8 + SZ_1K)
Why don't you keep the same macros and update them to deal with a
variable? It would make the patch so much easier to review.
Something along the lines of (untested):
diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
index 63cd0f2b8707..a1891840c66a 100644
--- a/drivers/irqchip/irq-gic-v3-its.c
+++ b/drivers/irqchip/irq-gic-v3-its.c
@@ -691,8 +691,10 @@ static struct irq_chip its_irq_chip = {
*/
#define IRQS_PER_CHUNK_SHIFT 5
#define IRQS_PER_CHUNK (1 << IRQS_PER_CHUNK_SHIFT)
+#define ITS_MAX_LPI_NRBITS (17) /* 128K LPIs */
static unsigned long *lpi_bitmap;
+static u32 lpi_nrbits;
static u32 lpi_chunks;
static DEFINE_SPINLOCK(lpi_lock);
@@ -706,9 +708,9 @@ static int its_chunk_to_lpi(int chunk)
return (chunk << IRQS_PER_CHUNK_SHIFT) + 8192;
}
-static int __init its_lpi_init(u32 id_bits)
+static int __init its_lpi_init(void)
{
- lpi_chunks = its_lpi_to_chunk(1UL << id_bits);
+ lpi_chunks = its_lpi_to_chunk(1UL << lpi_nrbits);
lpi_bitmap = kzalloc(BITS_TO_LONGS(lpi_chunks) * sizeof(long),
GFP_KERNEL);
@@ -793,13 +795,9 @@ static void its_lpi_free(struct event_lpi_map *map)
* deal with (one configuration byte per interrupt). PENDBASE has to
* be 64kB aligned (one bit per LPI, plus 8192 bits for SPI/PPI/SGI).
*/
-#define LPI_PROPBASE_SZ SZ_64K
-#define LPI_PENDBASE_SZ (LPI_PROPBASE_SZ / 8 + SZ_1K)
-
-/*
- * This is how many bits of ID we need, including the useless ones.
- */
-#define LPI_NRBITS ilog2(LPI_PROPBASE_SZ + SZ_8K)
+#define LPI_PROPBASE_SZ ALIGN(BIT(lpi_nrbits), SZ_64K)
+#define LPI_PENDBASE_SZ ALIGN(BIT(lpi_nrbits) / 8, SZ_64K)
+#define LPI_NRBITS lpi_nrbits
#define LPI_PROP_DEFAULT_PRIO 0xa0
@@ -807,6 +805,7 @@ static int __init its_alloc_lpi_tables(void)
{
phys_addr_t paddr;
+ lpi_nrbits = min_t(u32, gic_rdists->id_bits, ITS_MAX_LPI_NRBITS);
gic_rdists->prop_page = alloc_pages(GFP_NOWAIT,
get_order(LPI_PROPBASE_SZ));
if (!gic_rdists->prop_page) {
@@ -825,7 +824,7 @@ static int __init its_alloc_lpi_tables(void)
/* Make sure the GIC will observe the written configuration */
gic_flush_dcache_to_poc(page_address(gic_rdists->prop_page), LPI_PROPBASE_SZ);
- return 0;
+ return its_lpi_init();
}
static const char *its_base_type_string[] = {
@@ -1100,7 +1099,7 @@ static void its_cpu_init_lpis(void)
* hence the 'max(LPI_PENDBASE_SZ, SZ_64K)' below.
*/
pend_page = alloc_pages(GFP_NOWAIT | __GFP_ZERO,
- get_order(max(LPI_PENDBASE_SZ, SZ_64K)));
+ get_order(LPI_PENDBASE_SZ));
if (!pend_page) {
pr_err("Failed to allocate PENDBASE for CPU%d\n",
smp_processor_id());
@@ -1975,8 +1974,5 @@ int __init its_init(struct fwnode_handle *handle, struct rdists *rdists,
}
gic_rdists = rdists;
- its_alloc_lpi_tables();
- its_lpi_init(rdists->id_bits);
-
- return 0;
+ return its_alloc_lpi_tables();
}
Most of the patch is actually improving on the error handling...
Thanks,
M.
--
Jazz is not dead. It just smells funny...
Powered by blists - more mailing lists