From: Sebastian Andrzej Siewior <bigeasy-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
To: Joerg Roedel <joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org>
Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org,
tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org,
linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
Sebastian Andrzej Siewior
<bigeasy-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
Subject: [PATCH 09/10] iommu/amd: drop the lock while allocating new irq remap table
Date: Fri, 16 Mar 2018 21:18:35 +0100 [thread overview]
Message-ID: <20180316201836.7864-10-bigeasy@linutronix.de> (raw)
In-Reply-To: <20180316201836.7864-1-bigeasy-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
The irq_remap_table is allocated while the iommu_table_lock is held with
interrupts disabled. While this works it makes RT scream very loudly.
>From looking at the call sites, all callers are in the early device
initialisation (apic_bsp_setup(), pci_enable_device(),
pci_enable_msi()) so make sense to drop the lock which also enables
interrupts and try to allocate that memory with GFP_KERNEL instead
GFP_ATOMIC.
Since during the allocation the iommu_table_lock is dropped, we need to
recheck if table exists after the lock has been reacquired. I *think*
that it is impossible that the "devid" entry appears in irq_lookup_table
while the lock is dropped since the same device can only be probed once.
It is more likely that another device added an `alias' entry. However I
check for both cases, just to be sure.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
---
drivers/iommu/amd_iommu.c | 65 +++++++++++++++++++++++++++++++++--------------
1 file changed, 46 insertions(+), 19 deletions(-)
diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
index 7dd4c27a941d..0a7ca5e95288 100644
--- a/drivers/iommu/amd_iommu.c
+++ b/drivers/iommu/amd_iommu.c
@@ -3602,6 +3602,30 @@ static void set_dte_irq_entry(u16 devid, struct irq_remap_table *table)
amd_iommu_dev_table[devid].data[2] = dte;
}
+static struct irq_remap_table *alloc_irq_table(void)
+{
+ struct irq_remap_table *table;
+
+ table = kzalloc(sizeof(*table), GFP_ATOMIC);
+ if (!table)
+ return NULL;
+
+ table->table = kmem_cache_alloc(amd_iommu_irq_cache, GFP_KERNEL);
+ if (!table->table) {
+ kfree(table);
+ return NULL;
+ }
+ raw_spin_lock_init(&table->lock);
+
+ if (!AMD_IOMMU_GUEST_IR_GA(amd_iommu_guest_ir))
+ memset(table->table, 0,
+ MAX_IRQS_PER_TABLE * sizeof(u32));
+ else
+ memset(table->table, 0,
+ (MAX_IRQS_PER_TABLE * (sizeof(u64) * 2)));
+ return table;
+}
+
static void set_remap_table_entry(struct amd_iommu *iommu, u16 devid,
struct irq_remap_table *table)
{
@@ -3613,6 +3637,7 @@ static void set_remap_table_entry(struct amd_iommu *iommu, u16 devid,
static struct irq_remap_table *get_irq_table(u16 devid)
{
struct irq_remap_table *table = NULL;
+ struct irq_remap_table *new_table = NULL;
struct amd_iommu *iommu;
unsigned long flags;
u16 alias;
@@ -3631,42 +3656,44 @@ static struct irq_remap_table *get_irq_table(u16 devid)
table = irq_lookup_table[alias];
if (table) {
set_remap_table_entry(iommu, devid, table);
- goto out;
+ goto out_wait;
}
+ raw_spin_unlock_irqrestore(&iommu_table_lock, flags);
/* Nothing there yet, allocate new irq remapping table */
- table = kzalloc(sizeof(*table), GFP_ATOMIC);
- if (!table)
+ new_table = alloc_irq_table();
+ if (!new_table)
+ return NULL;
+
+ raw_spin_lock_irqsave(&iommu_table_lock, flags);
+
+ table = irq_lookup_table[devid];
+ if (table)
goto out_unlock;
- /* Initialize table spin-lock */
- raw_spin_lock_init(&table->lock);
-
- table->table = kmem_cache_alloc(amd_iommu_irq_cache, GFP_ATOMIC);
- if (!table->table) {
- kfree(table);
- table = NULL;
- goto out_unlock;
+ table = irq_lookup_table[alias];
+ if (table) {
+ set_remap_table_entry(iommu, devid, table);
+ goto out_wait;
}
- if (!AMD_IOMMU_GUEST_IR_GA(amd_iommu_guest_ir))
- memset(table->table, 0,
- MAX_IRQS_PER_TABLE * sizeof(u32));
- else
- memset(table->table, 0,
- (MAX_IRQS_PER_TABLE * (sizeof(u64) * 2)));
-
+ table = new_table;
+ new_table = NULL;
set_remap_table_entry(iommu, devid, table);
if (devid != alias)
set_remap_table_entry(iommu, alias, table);
-out:
+out_wait:
iommu_completion_wait(iommu);
out_unlock:
raw_spin_unlock_irqrestore(&iommu_table_lock, flags);
+ if (new_table) {
+ kmem_cache_free(amd_iommu_irq_cache, new_table->table);
+ kfree(new_table);
+ }
return table;
}
--
2.16.2
next prev parent reply other threads:[~2018-03-16 20:18 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-03-16 20:18 [PATCH 00/10 v2] iommu/amd: lock splitting & GFP_KERNEL allocation Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 05/10] Revert "iommu/amd: Avoid locking get_irq_table() from atomic context" Sebastian Andrzej Siewior
[not found] ` <20180316201836.7864-1-bigeasy-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
2018-03-16 20:18 ` [PATCH 01/10] iommu/amd: take into account that alloc_dev_data() may return NULL Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 02/10] iommu/amd: turn dev_data_list into a lock less list Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 03/10] iommu/amd: split domain id out of amd_iommu_devtable_lock Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 04/10] iommu/amd: split irq_lookup_table out of the amd_iommu_devtable_lock Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 06/10] iommu/amd: remove the special case from get_irq_table() Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 07/10] iommu/amd: use `table' instead `irt' as variable name in amd_iommu_update_ga() Sebastian Andrzej Siewior
2018-03-16 20:18 ` [PATCH 08/10] iommu/amd: factor out setting the remap table for a devid Sebastian Andrzej Siewior
2018-03-16 20:18 ` Sebastian Andrzej Siewior [this message]
2018-03-16 20:18 ` [PATCH 10/10] iommu/amd: make amd_iommu_devtable_lock a spin_lock Sebastian Andrzej Siewior
2018-03-17 19:49 ` [PATCH 00/10 v2] iommu/amd: lock splitting & GFP_KERNEL allocation Scott Wood
[not found] ` <1521316194.3722.74.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2018-03-17 21:10 ` Sebastian Andrzej Siewior
[not found] ` <20180317211013.rlou66s542ad4y2i-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
2018-03-17 21:43 ` Scott Wood
[not found] ` <1521323019.3722.98.camel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2018-03-19 12:15 ` Sebastian Andrzej Siewior
2018-03-19 22:49 ` Scott Wood
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180316201836.7864-10-bigeasy@linutronix.de \
--to=bigeasy-hfztesqfncyowbw4kg4ksq@public.gmane.org \
--cc=iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org \
--cc=joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org \
--cc=linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).