Skip to contentSkip to conversation

xhci features for usb-next

PATCH In review Mathias Nyman 2 participants 6 messages Jun 17, 2021

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover xhci features for usb-next

Hi Greg

A few small patches for usb-next.

There's one double free fix here as well that I normally would send to usb-linus, but we're late in the cycle and this issue should be rare. It has been there since 5.6 and requires system to be out of memory, so I thought it can be added this way.

Thanks
-Mathias

Mathias Nyman (3):
xhci: Remove unused defines for ERST_SIZE and ERST_ENTRIES
xhci: Add adaptive interrupt rate for isoch TRBs with XHCI_AVOID_BEI
quirk
xhci: handle failed buffer copy to URB sg list and fix a W=1 copiler
warning

Zhangjiantao (Kirin, nanjing) (1):
xhci: solve a double free problem while doing s4

 drivers/usb/host/xhci-mem.c  |  3 +++
 drivers/usb/host/xhci-ring.c |  7 ++++++-
 drivers/usb/host/xhci.c      |  9 +++++++--
 drivers/usb/host/xhci.h      | 11 +++++++----
 4 files changed, 23 insertions(+), 7 deletions(-)
Signature · 1 line
2.25.1
#1 lore raw

PATCH 1/4 xhci: Remove unused defines for ERST_SIZE and ERST_ENTRIES

We don't want those around confusing people.

ERST_NUM_SEGS is used both when allocating event ring segments, and when allocating entries in the event ring segment table (erst).

---
 drivers/usb/host/xhci.h | 4 ----
 1 file changed, 4 deletions(-)
drivers/usb/host/xhci.h +0 -4
index a1d5ffb7474d..85ba326806ab 100644
@@ -1663,10 +1663,6 @@ struct urb_priv {
  * meaning 64 ring segments.
  * Initial allocated size of the ERST, in number of entries */
 #define	ERST_NUM_SEGS	1
-/* Initial allocated size of the ERST, in number of entries */
-#define	ERST_SIZE	64
-/* Initial number of event segment rings allocated */
-#define	ERST_ENTRIES	1
 /* Poll every 60 seconds */
 #define	POLL_TIMEOUT	60
 /* Stop endpoint command timeout (secs) for URB cancellation watchdog timer */
Signature · 1 line
2.25.1
#2 lore raw

PATCH 2/4 xhci: Add adaptive interrupt rate for isoch TRBs with XHCI_AVOID_BEI quirk

In reply to Mathias Nyman

Save a bit of power by not interrupting so often by default if XHCI_AVOID_BEI quirk is set.

In normal cases the xhci driver will only generate an interrupt on the last isochronous TRB of an URB. In a common UVC webcam usecase there are 32 TRBs per URB.

if AVOID_BEI flag is set then xhci driver will force an interrupt every 8th isoc TRB to make sure the event ring doesn't get too full.

This is however way too frequent in common single webcam use cases, causing 1000 interrupts/sec and thus poor powermanagement performance.

Instead start with interrupting every 32 isoc TRB, and halve it in case event ring becomes half-full. Stop halving when reaching a rate of every 8th trb.

This is a one way solution. If interrupt rate is increased it will stay high until driver is reloaded. The highest rate is the same as the old default rate.

---
 drivers/usb/host/xhci-mem.c  | 2 ++
 drivers/usb/host/xhci-ring.c | 7 ++++++-
 drivers/usb/host/xhci.h      | 7 +++++++
 3 files changed, 15 insertions(+), 1 deletion(-)
drivers/usb/host/xhci-mem.c +2 -0
index f66815fe8482..2f6da35e7977 100644
@@ -2547,6 +2547,8 @@ int xhci_mem_init(struct xhci_hcd *xhci, gfp_t flags)
 	xhci_dbg_trace(xhci, trace_xhci_dbg_init,
 			"Wrote ERST address to ir_set 0.");
 
+	xhci->isoc_bei_interval = AVOID_BEI_INTERVAL_MAX;
+
 	/*
 	 * XXX: Might need to set the Interrupter Moderation Register to
 	 * something other than the default (~1ms minimum between interrupts).
drivers/usb/host/xhci-ring.c +6 -1
index 6acd2329e08d..8fea44bbc266 100644
@@ -3076,6 +3076,11 @@ irqreturn_t xhci_irq(struct usb_hcd *hcd)
 		if (event_loop++ < TRBS_PER_SEGMENT / 2)
 			continue;
 		xhci_update_erst_dequeue(xhci, event_ring_deq);
+
+		/* ring is half-full, force isoc trbs to interrupt more often */
+		if (xhci->isoc_bei_interval > AVOID_BEI_INTERVAL_MIN)
+			xhci->isoc_bei_interval = xhci->isoc_bei_interval / 2;
+
 		event_loop = 0;
 	}
 
@@ -3956,7 +3961,7 @@ static bool trb_block_event_intr(struct xhci_hcd *xhci, int num_tds, int i)
 	 * generate an event at least every 8th TD to clear the event ring
 	 */
 	if (i && xhci->quirks & XHCI_AVOID_BEI)
-		return !!(i % 8);
+		return !!(i % xhci->isoc_bei_interval);
 
 	return true;
 }
drivers/usb/host/xhci.h +7 -0
index 85ba326806ab..5ba01d5ccab8 100644
@@ -1526,6 +1526,12 @@ static inline const char *xhci_trb_type_string(u8 type)
 #define TRB_BUFF_LEN_UP_TO_BOUNDARY(addr)	(TRB_MAX_BUFF_SIZE - \
 					(addr & (TRB_MAX_BUFF_SIZE - 1)))
 #define MAX_SOFT_RETRY		3
+/*
+ * Limits of consecutive isoc trbs that can Block Event Interrupt (BEI) if
+ * XHCI_AVOID_BEI quirk is in use.
+ */
+#define AVOID_BEI_INTERVAL_MIN	8
+#define AVOID_BEI_INTERVAL_MAX	32
 
 struct xhci_segment {
 	union xhci_trb		*trbs;
@@ -1768,6 +1774,7 @@ struct xhci_hcd {
 	u8		isoc_threshold;
 	/* imod_interval in ns (I * 250ns) */
 	u32		imod_interval;
+	u32		isoc_bei_interval;
 	int		event_ring_max;
 	/* 4KB min, 128MB max */
 	int		page_size;
Signature · 1 line
2.25.1
#3 lore raw

PATCH 3/4 xhci: handle failed buffer copy to URB sg list and fix a W=1 copiler warning

In reply to Mathias Nyman

Set the urb->actual_length to bytes successfully copied in case all bytes weren't copied from a temporary buffer to the URB sg list. Also print a debug message

---
 drivers/usb/host/xhci.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/usb/host/xhci.c +7 -2
index 27283654ca08..9248ce8d09a4 100644
@@ -1361,12 +1361,17 @@ static void xhci_unmap_temp_buf(struct usb_hcd *hcd, struct urb *urb)
 				 urb->transfer_buffer_length,
 				 dir);
 
-	if (usb_urb_dir_in(urb))
+	if (usb_urb_dir_in(urb)) {
 		len = sg_pcopy_from_buffer(urb->sg, urb->num_sgs,
 					   urb->transfer_buffer,
 					   buf_len,
 					   0);
-
+		if (len != buf_len) {
+			xhci_dbg(hcd_to_xhci(hcd),
+				 "Copy from tmp buf to urb sg list failed\n");
+			urb->actual_length = len;
+		}
+	}
 	urb->transfer_flags &= ~URB_DMA_MAP_SINGLE;
 	kfree(urb->transfer_buffer);
 	urb->transfer_buffer = NULL;
Signature · 1 line
2.25.1
#4 lore raw

PATCH 4/4 xhci: solve a double free problem while doing s4

In reply to Mathias Nyman

From: "Zhangjiantao (Kirin, nanjing)" <[email protected]>

when system is doing s4, the process of xhci_resume may be as below:
1、xhci_mem_cleanup 2、xhci_init->xhci_mem_init->xhci_mem_cleanup(when memory is not enough). xhci_mem_cleanup will be executed twice when system is out of memory. xhci->port_caps is freed in xhci_mem_cleanup,but it isn't set to NULL. It will be freed twice when xhci_mem_cleanup is called the second time.

We got following bug when system resumes from s4:

kernel BUG at mm/slub.c:309!

Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
CPU: 0 PID: 5929 Tainted: G S   W   5.4.96-arm64-desktop #1
pc : __slab_free+0x5c/0x424
lr : kfree+0x30c/0x32c

Call trace:
 __slab_free+0x5c/0x424
 kfree+0x30c/0x32c
 xhci_mem_cleanup+0x394/0x3cc
 xhci_mem_init+0x9ac/0x1070
 xhci_init+0x8c/0x1d0
 xhci_resume+0x1cc/0x5fc
 xhci_plat_resume+0x64/0x70
 platform_pm_thaw+0x28/0x60
 dpm_run_callback+0x54/0x24c
 device_resume+0xd0/0x200
 async_resume+0x24/0x60
 async_run_entry_fn+0x44/0x110
 process_one_work+0x1f0/0x490
 worker_thread+0x5c/0x450
 kthread+0x158/0x160
 ret_from_fork+0x10/0x24

Original patch that caused this issue was backported to 4.4 stable, so this should be backported to 4.4 stabe as well.

Fixes cf0ee7c60c89 ("xhci: Fix memory leak when caching protocol extended capability PSI tables - take 2")
---
 drivers/usb/host/xhci-mem.c | 1 +
 1 file changed, 1 insertion(+)
drivers/usb/host/xhci-mem.c +1 -0
index 2f6da35e7977..0e312066c5c6 100644
@@ -1924,6 +1924,7 @@ void xhci_mem_cleanup(struct xhci_hcd *xhci)
 	xhci->hw_ports = NULL;
 	xhci->rh_bw = NULL;
 	xhci->ext_caps = NULL;
+	xhci->port_caps = NULL;
 
 	xhci->page_size = 0;
 	xhci->page_shift = 0;
Signature · 1 line
2.25.1
maintainer
#5 lore raw
In reply to Mathias Nyman
Quoted 8 lines · Mathias Nyman

Hi Greg

A few small patches for usb-next.

There's one double free fix here as well that I normally would send to usb-linus, but we're late in the cycle and this issue should be rare. It has been there since 5.6 and requires system to be out of memory, so I thought it can be added this way.

Yes, that's fine, all now applied, thanks.

greg k-h