improve error handling on chipidea/udc.c and f_midi.c
Subsystems (2)
cover improve error handling on chipidea/udc.c and f_midi.c
From: "Felipe F. Tonello" <[email protected]>
Felipe F. Tonello (2):
usb: chipidea: udc: improve error handling on ep_queue
usb: gadget: f_midi: check for error on usb_ep_queue
drivers/usb/chipidea/udc.c | 26 +++++++++++++++++++------- drivers/usb/gadget/function/f_midi.c | 10 ++++++++-- 2 files changed, 27 insertions(+), 9 deletions(-)
Signature · 1 line
2.1.4
PATCH 1/2 usb: chipidea: udc: improve error handling on ep_queue
From: "Felipe F. Tonello" <[email protected]>
_ep_queue() didn't check for errors when using add_td_to_list() which can fail if dma_pool_alloc fails, thus causing a kernel panic when lastnode->ptr is NULL.
--- drivers/usb/chipidea/udc.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-)
drivers/usb/chipidea/udc.c +19 -7
@@ -404,9 +404,9 @@ static inline u8 _usb_addr(struct ci_hw_ep *ep) }/** -* _hardware_queue: configures a request at hardware level -* @gadget: gadget +* _hardware_enqueue: configures a request at hardware level * @hwep: endpoint +* @hwreq: request * * This function returns an error code */
@@ -435,19 +435,27 @@ static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq)if (hwreq->req.dma %PAGE_SIZE ) pages--; -if (rest ==0 ) -add_td_to_list (hwep, hwreq,0 ); +if (rest ==0 ) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + }while (rest >0 ) {unsigned count =min (hwreq->req.length - hwreq->req.actual, (unsigned )(pages *CI_HDRC_PAGE_SIZE )); -add_td_to_list (hwep, hwreq, count); + ret =add_td_to_list (hwep, hwreq, count); +if (ret <0 ) + goto done; rest -= count; }if (hwreq->req.zero && hwreq->req.length - && (hwreq->req.length % hwep->ep.maxpacket ==0 )) -add_td_to_list (hwep, hwreq,0 ); + && (hwreq->req.length % hwep->ep.maxpacket ==0 )) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + } firstnode = list_first_entry(&hwreq->tds, struct td_node, td);
@@ -750,8 +758,12 @@ static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req)/** * _ep_queue: queues (submits) an I/O request to an endpoint +* @ep: endpoint +* @req: request +* @gfp_flags: GFP flags (not used) * * Caller must hold lock +* This function returns an error code */ static int _ep_queue (struct usb_ep *ep,struct usb_request *req,gfp_t __maybe_unused gfp_flags)
Signature · 1 line
2.1.4
Re: [PATCH 1/2] usb: chipidea: udc: improve error handling on ep_queue
Hi,
Quoted 7 lines · On Fri, Sep 18, 2015 at 06:12:40PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
_ep_queue()didn't check for errors when usingadd_td_to_list()which can fail ifdma_pool_allocfails, thus causing a kernel panic whenlastnode->ptris NULL.Signed-off-by Felipe F. Tonello <[email protected]>
this can still be split down further.
Quoted 9 lines
--- drivers/usb/chipidea/udc.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-)drivers/usb/chipidea/udc.c +0
-0
drivers/usb/chipidea/udc.c static inline u8 _usb_addr(struct ci_hw_ep *ep) · line 409
Quoted 8 lines
drivers/usb/chipidea/udc.c +2
-2@@ -404,9 +404,9 @@ static inline u8 _usb_addr(struct ci_hw_ep *ep) }/** -* _hardware_queue: configures a request at hardware level -* @gadget: gadget +* _hardware_enqueue: configures a request at hardware level * @hwep: endpoint +* @hwreq: request
this is a cleanup and you shouldn't have a fix depending on a cleanup. Fixes are merged during the -rc cycle, while cleanups will be deferred to the following merge window.
Quoted 4 lines
drivers/usb/chipidea/udc.c +0
-0* * This function returns an error code */
drivers/usb/chipidea/udc.c static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq) · line 442
Quoted 10 lines
drivers/usb/chipidea/udc.c +5
-2@@ -435,19 +435,27 @@ static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq)if (hwreq->req.dma %PAGE_SIZE ) pages--; -if (rest ==0 ) -add_td_to_list (hwep, hwreq,0 ); +if (rest ==0 ) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + }
this is your fix.
drivers/usb/chipidea/udc.c static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq) · line 449
Quoted 8 lines
drivers/usb/chipidea/udc.c +3
-1while (rest >0 ) {unsigned count =min (hwreq->req.length - hwreq->req.actual, (unsigned )(pages *CI_HDRC_PAGE_SIZE )); -add_td_to_list (hwep, hwreq, count); + ret =add_td_to_list (hwep, hwreq, count); +if (ret <0 ) + goto done;
and this
drivers/usb/chipidea/udc.c static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq) · line 459
Quoted 12 lines
drivers/usb/chipidea/udc.c +5
-2rest -= count; }if (hwreq->req.zero && hwreq->req.length - && (hwreq->req.length % hwep->ep.maxpacket ==0 )) -add_td_to_list (hwep, hwreq,0 ); + && (hwreq->req.length % hwep->ep.maxpacket ==0 )) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + }
and this.
drivers/usb/chipidea/udc.c static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req) · line 763
Quoted 9 lines
drivers/usb/chipidea/udc.c +3
-0firstnode = list_first_entry(&hwreq->tds, struct td_node, td);@@ -750,8 +758,12 @@ static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req)/** * _ep_queue: queues (submits) an I/O request to an endpoint +* @ep: endpoint +* @req: request +* @gfp_flags: GFP flags (not used)
cleanup
drivers/usb/chipidea/udc.c static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req) · line 766
Quoted 3 lines
drivers/usb/chipidea/udc.c +1
-0* * Caller must hold lock +* This function returns an error code
somewhat pointless, but could come with the cleanup, no strong feelings.
Quoted 6 lines
drivers/usb/chipidea/udc.c +0
-0*/ static int _ep_queue (struct usb_ep *ep,struct usb_request *req,gfp_t __maybe_unused gfp_flags)Signature · 1 line
2.1.4
Signature · 1 line
balbi
Re: [PATCH 1/2] usb: chipidea: udc: improve error handling on ep_queue
Hi Felipe,
Quoted 24 lines · Felipe Balbi
Hi,
Quoted 7 lines · On Fri, Sep 18, 2015 at 06:12:40PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
_ep_queue()didn't check for errors when usingadd_td_to_list()which can fail ifdma_pool_allocfails, thus causing a kernel panic whenlastnode->ptris NULL.Signed-off-by Felipe F. Tonello <[email protected]>this can still be split down further.
Quoted 11 lines
--- drivers/usb/chipidea/udc.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-)drivers/usb/chipidea/udc.c +0
-0@@ -404,9 +404,9 @@ static inline u8 _usb_addr(struct ci_hw_ep *ep) }
Quoted 10 lines
drivers/usb/chipidea/udc.c static inline u8 _usb_addr(struct ci_hw_ep *ep) · line 409
Quoted 6 lines
drivers/usb/chipidea/udc.c +2
-2/** -* _hardware_queue: configures a request at hardware level -* @gadget: gadget +* _hardware_enqueue: configures a request at hardware level * @hwep: endpoint +* @hwreq: request this is a cleanup and you shouldn't have a fix depending on a cleanup. Fixes are merged during the -rc cycle, while cleanups will be deferred to the following merge window.
Got it.
Quoted 73 lines
Quoted 5 lines
drivers/usb/chipidea/udc.c +0
-0* * This function returns an error code */ drivers/usb/chipidea/udc.c static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq) · line 442
Quoted 10 lines
drivers/usb/chipidea/udc.c +5
-2@@ -435,19 +435,27 @@ static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq)if (hwreq->req.dma %PAGE_SIZE ) pages--; -if (rest ==0 ) -add_td_to_list (hwep, hwreq,0 ); +if (rest ==0 ) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + }this is your fix.
drivers/usb/chipidea/udc.c static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq) · line 449
Quoted 8 lines
drivers/usb/chipidea/udc.c +3
-1while (rest >0 ) {unsigned count =min (hwreq->req.length - hwreq->req.actual, (unsigned )(pages *CI_HDRC_PAGE_SIZE )); -add_td_to_list (hwep, hwreq, count); + ret =add_td_to_list (hwep, hwreq, count); +if (ret <0 ) + goto done;and this
drivers/usb/chipidea/udc.c static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq) · line 459
Quoted 12 lines
drivers/usb/chipidea/udc.c +5
-2rest -= count; }if (hwreq->req.zero && hwreq->req.length - && (hwreq->req.length % hwep->ep.maxpacket ==0 )) -add_td_to_list (hwep, hwreq,0 ); + && (hwreq->req.length % hwep->ep.maxpacket ==0 )) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + }and this.
drivers/usb/chipidea/udc.c static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req) · line 763
Quoted 9 lines
drivers/usb/chipidea/udc.c +3
-0firstnode = list_first_entry(&hwreq->tds, struct td_node, td);@@ -750,8 +758,12 @@ static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req)/** * _ep_queue: queues (submits) an I/O request to an endpoint +* @ep: endpoint +* @req: request +* @gfp_flags: GFP flags (not used) cleanup
drivers/usb/chipidea/udc.c static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req) · line 766
Quoted 3 lines
drivers/usb/chipidea/udc.c +1
-0* * Caller must hold lock +* This function returns an error code somewhat pointless, but could come with the cleanup, no strong feelings.
Quoted 6 lines
drivers/usb/chipidea/udc.c +0
-1*/ static int _ep_queue (struct usb_ep *ep,struct usb_request *req,gfp_t __maybe_unused gfp_flags) --2.1.4
--
balbi
Re: [PATCH 1/2] usb: chipidea: udc: improve error handling on ep_queue
In reply to [email protected]Quoted 4 lines · On Fri, Sep 18, 2015 at 06:12:40PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
_ep_queue()didn't check for errors when usingadd_td_to_list()which can fail ifdma_pool_allocfails, thus causing a kernel
Would you find the root cause why dma_pool_alloc fails?
Quoted 72 lines
panic when
lastnode->ptris NULL.Signed-off-by Felipe F. Tonello <[email protected]>--- drivers/usb/chipidea/udc.c | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-)drivers/usb/chipidea/udc.c +19
-7@@ -404,9 +404,9 @@ static inline u8 _usb_addr(struct ci_hw_ep *ep) }/** -* _hardware_queue: configures a request at hardware level -* @gadget: gadget +* _hardware_enqueue: configures a request at hardware level * @hwep: endpoint +* @hwreq: request * * This function returns an error code */ @@ -435,19 +435,27 @@ static int _hardware_enqueue(struct ci_hw_ep *hwep, struct ci_hw_req *hwreq)if (hwreq->req.dma %PAGE_SIZE ) pages--; -if (rest ==0 ) -add_td_to_list (hwep, hwreq,0 ); +if (rest ==0 ) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + }while (rest >0 ) {unsigned count =min (hwreq->req.length - hwreq->req.actual, (unsigned )(pages *CI_HDRC_PAGE_SIZE )); -add_td_to_list (hwep, hwreq, count); + ret =add_td_to_list (hwep, hwreq, count); +if (ret <0 ) + goto done; rest -= count; }if (hwreq->req.zero && hwreq->req.length - && (hwreq->req.length % hwep->ep.maxpacket ==0 )) -add_td_to_list (hwep, hwreq,0 ); + && (hwreq->req.length % hwep->ep.maxpacket ==0 )) { + ret =add_td_to_list (hwep, hwreq,0 ); +if (ret <0 ) + goto done; + } firstnode = list_first_entry(&hwreq->tds, struct td_node, td);@@ -750,8 +758,12 @@ static void isr_get_status_complete(struct usb_ep *ep, struct usb_request *req)/** * _ep_queue: queues (submits) an I/O request to an endpoint +* @ep: endpoint +* @req: request +* @gfp_flags: GFP flags (not used) * * Caller must hold lock +* This function returns an error code */ static int _ep_queue (struct usb_ep *ep,struct usb_request *req,gfp_t __maybe_unused gfp_flags)Signature · 1 line
2.1.4
Signature · 2 lines
Best Regards, Peter Chen
Re: [PATCH 1/2] usb: chipidea: udc: improve error handling on ep_queue
Hi Peter,
Quoted 7 lines · Peter Chen
Quoted 4 lines · On Fri, Sep 18, 2015 at 06:12:40PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
_ep_queue()didn't check for errors when usingadd_td_to_list()which can fail ifdma_pool_allocfails, thus causing a kernelWould you find the root cause why
dma_pool_allocfails?
Partially. For some reason the udc_irq() is not been called after a f_midi_transmit(). I am looking into that.
Felipe
PATCH 2/2 usb: gadget: f_midi: check for error on usb_ep_queue
In reply to [email protected]From: "Felipe F. Tonello" <[email protected]>
f_midi is not checking weather the is an error on usb_ep_queue request, ignoring potential problems, such as memory leaks.
--- drivers/usb/gadget/function/f_midi.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/usb/gadget/function/f_midi.c +8 -2
@@ -543,8 +543,14 @@ static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) } } -if (req->length >0 ) -usb_ep_queue (ep, req,GFP_ATOMIC ); +if (req->length >0 ) { +int err; + + err =usb_ep_queue (ep, req,GFP_ATOMIC ); +if (err <0 ) +ERROR (midi,"%s queue req: %d\n" , + midi->out_ep->name, err); + }else free_ep_req (ep, req); }
Signature · 1 line
2.1.4
Re: [PATCH 2/2] usb: gadget: f_midi: check for error on usb_ep_queue
Quoted 19 lines · On Fri, Sep 18, 2015 at 06:12:41PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
f_midiis not checking weather the is an error onusb_ep_queuerequest, ignoring potential problems, such as memory leaks.Signed-off-by Felipe F. Tonello <[email protected]>--- drivers/usb/gadget/function/f_midi.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)drivers/usb/gadget/function/f_midi.c +0
-1@@ -543,8 +543,14 @@ static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) } } -if (req->length >0 )
drivers/usb/gadget/function/f_midi.c static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) · line 554
Quoted 10 lines
drivers/usb/gadget/function/f_midi.c +8
-1-usb_ep_queue (ep, req,GFP_ATOMIC ); +if (req->length >0 ) { +int err; + + err =usb_ep_queue (ep, req,GFP_ATOMIC ); +if (err <0 ) +ERROR (midi,"%s queue req: %d\n" , + midi->out_ep->name, err); + }else
yeah, cool, but you need to fix the style here. This else needs to be after the curly brace and you need to curly brace to the else branch too.
Quoted 5 lines
drivers/usb/gadget/function/f_midi.c +0
-0free_ep_req (ep, req); }Signature · 1 line
2.1.4
Signature · 1 line
balbi
Re: [PATCH 2/2] usb: gadget: f_midi: check for error on usb_ep_queue
In reply to [email protected]Hello.
Quoted 19 lines · On 9/18/2015 8:12 PM, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
f_midiis not checking weather the is an error onusb_ep_queuerequest, ignoring potential problems, such as memory leaks.Signed-off-by Felipe F. Tonello <[email protected]>--- drivers/usb/gadget/function/f_midi.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)drivers/usb/gadget/function/f_midi.c +0
-1@@ -543,8 +543,14 @@ static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) } } -if (req->length >0 )
drivers/usb/gadget/function/f_midi.c static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) · line 554
Quoted 10 lines
drivers/usb/gadget/function/f_midi.c +8
-1-usb_ep_queue (ep, req,GFP_ATOMIC ); +if (req->length >0 ) { +int err; + + err =usb_ep_queue (ep, req,GFP_ATOMIC ); +if (err <0 ) +ERROR (midi,"%s queue req: %d\n" , + midi->out_ep->name, err); + }else
} else {drivers/usb/gadget/function/f_midi.c static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) · line 555
Quoted 1 line
drivers/usb/gadget/function/f_midi.c +0
-0free_ep_req (ep, req);
}
drivers/usb/gadget/function/f_midi.c static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) · line 556
Quoted 1 line
drivers/usb/gadget/function/f_midi.c +0
-0}
The lines added above show the proper kernel CodingStyle. } should be on
the same line as *else* and {} should be used in all branches of the *if* statement if at least one branch has them.
MBR, Sergei
Re: [PATCH 2/2] usb: gadget: f_midi: check for error on usb_ep_queue
In reply to [email protected]Quoted 3 lines · On Fri, Sep 18, 2015 at 06:12:41PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
f_midiis not checking weather the is an error onusb_ep_queue
%s/weather/whether
%s/the/there
Quoted 31 lines
request, ignoring potential problems, such as memory leaks.
Signed-off-by Felipe F. Tonello <[email protected]>--- drivers/usb/gadget/function/f_midi.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)drivers/usb/gadget/function/f_midi.c +8
-2@@ -543,8 +543,14 @@ static void f_midi_transmit(struct f_midi *midi, struct usb_request *req) } } -if (req->length >0 ) -usb_ep_queue (ep, req,GFP_ATOMIC ); +if (req->length >0 ) { +int err; + + err =usb_ep_queue (ep, req,GFP_ATOMIC ); +if (err <0 ) +ERROR (midi,"%s queue req: %d\n" , + midi->out_ep->name, err); + }else free_ep_req (ep, req); }Signature · 1 line
2.1.4
Signature · 2 lines
Best Regards, Peter Chen
Re: [PATCH 2/2] usb: gadget: f_midi: check for error on usb_ep_queue
Hi Chen,
Quoted 7 lines · Peter Chen
Quoted 3 lines · On Fri, Sep 18, 2015 at 06:12:41PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
f_midiis not checking weather the is an error onusb_ep_queue%s/weather/whether
%s/the/there
I fixed it on v3. Did you receive it?
Felipe
Re: [PATCH 2/2] usb: gadget: f_midi: check for error on usb_ep_queue
Quoted 4 lines · Felipe Tonello
Hi Chen,
Quoted 1 line · Peter Chen
On Fri, Sep 18, 2015 at 06:12:41PM +0100, [email protected] wrote:
Quoted 9 lines
Quoted 6 lines
Quoted 3 lines
From: "Felipe F. Tonello" <[email protected]>
f_midiis not checking weather the is an error onusb_ep_queue%s/weather/whether
%s/the/thereI fixed it on v3. Did you receive it?
oh, get it, You may add "--subject-prefix="PATCH v3" when you format patch, in that case, the reader can see "PATCH v3 2/2" at the subject, and know it is the third version.
Your two changes for chipidea are ok, I will queue them.
Signature · 2 lines
Best Regards, Peter Chen
Re: [PATCH 2/2] usb: gadget: f_midi: check for error on usb_ep_queue
Hi Peter,
Quoted 12 lines · Peter Chen
Quoted 11 lines · Felipe Tonello
Hi Chen,
Quoted 7 lines · Peter Chen
Quoted 3 lines · On Fri, Sep 18, 2015 at 06:12:41PM +0100, [email protected] wrote:
From: "Felipe F. Tonello" <[email protected]>
f_midiis not checking weather the is an error onusb_ep_queue%s/weather/whether
%s/the/there
Quoted 8 lines
Quoted 2 lines
I fixed it on v3. Did you receive it?
oh, get it, You may add "--subject-prefix="PATCH v3" when you format patch, in that case, the reader can see "PATCH v3 2/2" at the subject, and know it is the third version.
Your two changes for chipidea are ok, I will queue them.
Thanks. Sorry about that, I thought I did it.
Felipe
2015-09-18 14:12 GMT-03:00 <[email protected]>:
Quoted 1 line
From: "Felipe F. Tonello" <[email protected]>
Signed-off-by ?
Albino
Quoted 4 lines · Albino B Neto
2015-09-18 14:12 GMT-03:00 <[email protected]>:
Quoted 1 line
From: "Felipe F. Tonello" <[email protected]>
Signed-off-by ?
this is not a patch
Signature · 1 line
balbi