All of lore.kernel.org
 help / color / mirror / Atom feed
From: Arend van Spriel <arend.vanspriel@broadcom.com>
To: Wright Feng <wright.feng@cypress.com>,
	franky.lin@broadcom.com, hante.meuleman@broadcom.com,
	kvalo@codeaurora.org, chi-hsien.lin@cypress.com
Cc: linux-wireless@vger.kernel.org, brcm80211-dev-list.pdl@broadcom.com
Subject: Re: [PATCH] brcmfmac: transfer firmware error to Linux error code in fwil
Date: Thu, 23 Nov 2017 11:41:22 +0100	[thread overview]
Message-ID: <5A16A5D2.8020600@broadcom.com> (raw)
In-Reply-To: <1511341876-10198-1-git-send-email-wright.feng@cypress.com>

On 11/22/2017 10:11 AM, Wright Feng wrote:
> fil_cmd_data_set and fil_cmd_data_get return proprietary error code
> when getting error from firmware layer. The vendor tools or utilities
> that uses libnl may stuck in some commands when wl is down.
> For example, issue "scan" command after issuing "down" command, the
> "scan" command will be the blocking call and stuck as no response from
> libnl. It is caused by that firmware returns BCME_NOTUP(-4) when wl
> is down, but the -4 is -EINTR in Linux kernel, so libnl catches the
> error and not passes to upper layer.
> Because of that, the fwil should return Linux error code instead of
> the proprietary error code, and the tools or utilities should get the
> real firmware error code by command "bcmerror" or "bcmerrorstr" after
> receiving the error from libnl.

Hi Wright,

I recall you had a discussion (on the wireless list?) with Franky about 
this, but I did not chime in at that time. I am not a proponent for 
mapping our firmware errors to linux errors. It is tempting to do so 
because there is clear overlap for some, but still it is quirky and has 
no value in my opinion. This issue was already addressed in the past 
although not specifically for fixing the -EINTR case. I did some digging 
in the git history. Here was my original change:

commit 9c6476668025e76a8365be783f2649fe3259b91c
Author: Arend van Spriel <arend@broadcom.com>
Date:   Tue Oct 28 14:56:09 2014 +0100

     brcmfmac: do not use firmware error code in driver

     Passing the firmware error codes up the driver may be mapped to
     linux error numbers which may impact proper fault analysis. So
     better pass up a generic failure code, ie. -EBADE and only show
     firmware error code in FIL debug message.

     Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
     Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
     Reviewed-by: Daniel (Deognyoun) Kim <dekim@broadcom.com>
     Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
     Signed-off-by: Arend van Spriel <arend@broadcom.com>
     Signed-off-by: John W. Linville <linville@tuxdriver.com>

The idea here was to have a single error code and have a debug message 
showing the firmware error string. This single error code could be used 
by vendor tools (not really an upstream concern) to query the driver for 
the firmware error code through a vendor specific command.

Now what happened is that Hante while inspecting logs with only error 
messages observed -EBADE and wanted to know firmware error. So he made 
the following patch a year later:

commit 21000b3f3da45a5a33de3815f3c2b3584102960e
Author: Hante Meuleman <meuleman@broadcom.com>
Date:   Wed Nov 25 11:32:38 2015 +0100

     brcmfmac: Return actual error by fwil.

     FWIL is always mapping back errors to EBADE. This is not very
     conventient when trying to understand problems by reading logs.
     Some callers print the error code, but that is quite useless
     when the exact error code is not returned. It also makes it
     impossible to differentiate based on error code. This patch
     changes the return of EBADE into the actual error code.

     Reviewed-by: Arend Van Spriel <arend@broadcom.com>
     Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
     Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
     Signed-off-by: Hante Meuleman <meuleman@broadcom.com>
     Signed-off-by: Arend van Spriel <arend@broadcom.com>
     Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

However, the consequence of this was that we ended up returning firmware 
error codes into the kernel domain and user-space.

Below the change I came up with. Let me know what you think.

Regards,
Arend
---8<--------------------------------------------------------
commit e340892a4bb5a74248b19504bb972497d38a4a03
Author: Arend van Spriel <arend.vanspriel@broadcom.com>
Date:   Thu Nov 23 11:13:41 2017 +0100

     brcmfmac: separate firmware errors from i/o errors

     When using the firmware api it can fail simply because firmware does
     not like the request or it fails due to issues in the host interface.
     Currently, there is only a single error code which is confusing. So
     adding a parameter to pass the firmware error separately and in case
     of a firmware error always return -EBADE to user-space.

     Reviewed-by: Hante Meuleman <hante.meuleman@broadcom.com>
     Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
     Reviewed-by: Franky Lin <franky.lin@broadcom.com>
     Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/bcdc.c 
b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/bcdc.c
index 9f2d0b0..353ed28 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/bcdc.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/bcdc.c
@@ -165,7 +165,7 @@ static int brcmf_proto_bcdc_cmplt(struct brcmf_pub 
*drvr, u32 id, u32 len)

  static int
  brcmf_proto_bcdc_query_dcmd(struct brcmf_pub *drvr, int ifidx, uint cmd,
-			    void *buf, uint len)
+			    void *buf, uint len, int *fwerr)
  {
  	struct brcmf_bcdc *bcdc = (struct brcmf_bcdc *)drvr->proto->pd;
  	struct brcmf_proto_bcdc_dcmd *msg = &bcdc->msg;
@@ -175,6 +175,7 @@ static int brcmf_proto_bcdc_cmplt(struct brcmf_pub 
*drvr, u32 id, u32 len)

  	brcmf_dbg(BCDC, "Enter, cmd %d len %d\n", cmd, len);

+	*fwerr = 0;
  	ret = brcmf_proto_bcdc_msg(drvr, ifidx, cmd, buf, len, false);
  	if (ret < 0) {
  		brcmf_err("brcmf_proto_bcdc_msg failed w/status %d\n",
@@ -211,17 +212,18 @@ static int brcmf_proto_bcdc_cmplt(struct brcmf_pub 
*drvr, u32 id, u32 len)
  		memcpy(buf, info, len);
  	}

+	ret = 0;
+
  	/* Check the ERROR flag */
  	if (flags & BCDC_DCMD_ERROR)
-		ret = le32_to_cpu(msg->status);
-
+		*fwerr = le32_to_cpu(msg->status);
  done:
  	return ret;
  }

  static int
  brcmf_proto_bcdc_set_dcmd(struct brcmf_pub *drvr, int ifidx, uint cmd,
-			  void *buf, uint len)
+			  void *buf, uint len, int *fwerr)
  {
  	struct brcmf_bcdc *bcdc = (struct brcmf_bcdc *)drvr->proto->pd;
  	struct brcmf_proto_bcdc_dcmd *msg = &bcdc->msg;
@@ -230,6 +232,7 @@ static int brcmf_proto_bcdc_cmplt(struct brcmf_pub 
*drvr, u32 id, u32 len)

  	brcmf_dbg(BCDC, "Enter, cmd %d len %d\n", cmd, len);

+	*fwerr = 0;
  	ret = brcmf_proto_bcdc_msg(drvr, ifidx, cmd, buf, len, true);
  	if (ret < 0)
  		goto done;
@@ -251,7 +254,7 @@ static int brcmf_proto_bcdc_cmplt(struct brcmf_pub 
*drvr, u32 id, u32 len)

  	/* Check the ERROR flag */
  	if (flags & BCDC_DCMD_ERROR)
-		ret = le32_to_cpu(msg->status);
+		*fwerr = le32_to_cpu(msg->status);

  done:
  	return ret;
diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwil.c 
b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwil.c
index f6a2df9..688f256 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwil.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwil.c
@@ -107,7 +107,7 @@ static const char *brcmf_fil_get_errstr(u32 err)
  brcmf_fil_cmd_data(struct brcmf_if *ifp, u32 cmd, void *data, u32 len, 
bool set)
  {
  	struct brcmf_pub *drvr = ifp->drvr;
-	s32 err;
+	s32 err, fwerr;

  	if (drvr->bus_if->state != BRCMF_BUS_UP) {
  		brcmf_err("bus is down. we have nothing to do.\n");
@@ -117,16 +117,17 @@ static const char *brcmf_fil_get_errstr(u32 err)
  	if (data != NULL)
  		len = min_t(uint, len, BRCMF_DCMD_MAXLEN);
  	if (set)
-		err = brcmf_proto_set_dcmd(drvr, ifp->ifidx, cmd, data, len);
+		err = brcmf_proto_set_dcmd(drvr, ifp->ifidx, cmd,
+					   data, len, &fwerr);
  	else
-		err = brcmf_proto_query_dcmd(drvr, ifp->ifidx, cmd, data, len);
-
-	if (err >= 0)
-		return 0;
-
-	brcmf_dbg(FIL, "Failed: %s (%d)\n",
-		  brcmf_fil_get_errstr((u32)(-err)), err);
+		err = brcmf_proto_query_dcmd(drvr, ifp->ifidx, cmd,
+					     data, len, &fwerr);

+	if (fwerr < 0) {
+		brcmf_dbg(FIL, "Firmware error: %s (%d)\n",
+			  brcmf_fil_get_errstr((u32)(-fwerr)), fwerr);
+		err = -EBADE;
+	}
  	return err;
  }

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/msgbuf.c 
b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/msgbuf.c
index d2c834c..e212a79 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/msgbuf.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/msgbuf.c
@@ -477,7 +477,7 @@ static void brcmf_msgbuf_ioctl_resp_wake(struct 
brcmf_msgbuf *msgbuf)


  static int brcmf_msgbuf_query_dcmd(struct brcmf_pub *drvr, int ifidx,
-				   uint cmd, void *buf, uint len)
+				   uint cmd, void *buf, uint len, int *fwerr)
  {
  	struct brcmf_msgbuf *msgbuf = (struct brcmf_msgbuf *)drvr->proto->pd;
  	struct sk_buff *skb = NULL;
@@ -485,6 +485,7 @@ static int brcmf_msgbuf_query_dcmd(struct brcmf_pub 
*drvr, int ifidx,
  	int err;

  	brcmf_dbg(MSGBUF, "ifidx=%d, cmd=%d, len=%d\n", ifidx, cmd, len);
+	*fwerr = 0;
  	msgbuf->ctl_completed = false;
  	err = brcmf_msgbuf_tx_ioctl(drvr, ifidx, cmd, buf, len);
  	if (err)
@@ -508,14 +509,15 @@ static int brcmf_msgbuf_query_dcmd(struct 
brcmf_pub *drvr, int ifidx,
  	}
  	brcmu_pkt_buf_free_skb(skb);

-	return msgbuf->ioctl_resp_status;
+	*fwerr = msgbuf->ioctl_resp_status;
+	return 0;
  }


  static int brcmf_msgbuf_set_dcmd(struct brcmf_pub *drvr, int ifidx,
-				 uint cmd, void *buf, uint len)
+				 uint cmd, void *buf, uint len, int *fwerr)
  {
-	return brcmf_msgbuf_query_dcmd(drvr, ifidx, cmd, buf, len);
+	return brcmf_msgbuf_query_dcmd(drvr, ifidx, cmd, buf, len, fwerr);
  }


diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/proto.h 
b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/proto.h
index 2404f8a..8a8e08f 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/proto.h
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/proto.h
@@ -30,9 +30,9 @@ struct brcmf_proto {
  	int (*hdrpull)(struct brcmf_pub *drvr, bool do_fws,
  		       struct sk_buff *skb, struct brcmf_if **ifp);
  	int (*query_dcmd)(struct brcmf_pub *drvr, int ifidx, uint cmd,
-			  void *buf, uint len);
+			  void *buf, uint len, int *fwerr);
  	int (*set_dcmd)(struct brcmf_pub *drvr, int ifidx, uint cmd, void *buf,
-			uint len);
+			uint len, int *fwerr);
  	int (*tx_queue_data)(struct brcmf_pub *drvr, int ifidx,
  			     struct sk_buff *skb);
  	int (*txdata)(struct brcmf_pub *drvr, int ifidx, u8 offset,
@@ -71,14 +71,16 @@ static inline int brcmf_proto_hdrpull(struct 
brcmf_pub *drvr, bool do_fws,
  	return drvr->proto->hdrpull(drvr, do_fws, skb, ifp);
  }
  static inline int brcmf_proto_query_dcmd(struct brcmf_pub *drvr, int 
ifidx,
-					 uint cmd, void *buf, uint len)
+					 uint cmd, void *buf, uint len,
+					 int *fwerr)
  {
-	return drvr->proto->query_dcmd(drvr, ifidx, cmd, buf, len);
+	return drvr->proto->query_dcmd(drvr, ifidx, cmd, buf, len,fwerr);
  }
  static inline int brcmf_proto_set_dcmd(struct brcmf_pub *drvr, int ifidx,
-				       uint cmd, void *buf, uint len)
+				       uint cmd, void *buf, uint len,
+				       int *fwerr)
  {
-	return drvr->proto->set_dcmd(drvr, ifidx, cmd, buf, len);
+	return drvr->proto->set_dcmd(drvr, ifidx, cmd, buf, len, fwerr);
  }

  static inline int brcmf_proto_tx_queue_data(struct brcmf_pub *drvr, 
int ifidx,

  reply	other threads:[~2017-11-23 10:41 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-11-22  9:11 [PATCH] brcmfmac: transfer firmware error to Linux error code in fwil Wright Feng
2017-11-23 10:41 ` Arend van Spriel [this message]
2017-11-24  6:35   ` Wright Feng
2017-11-24 10:22     ` Arend van Spriel

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=5A16A5D2.8020600@broadcom.com \
    --to=arend.vanspriel@broadcom.com \
    --cc=brcm80211-dev-list.pdl@broadcom.com \
    --cc=chi-hsien.lin@cypress.com \
    --cc=franky.lin@broadcom.com \
    --cc=hante.meuleman@broadcom.com \
    --cc=kvalo@codeaurora.org \
    --cc=linux-wireless@vger.kernel.org \
    --cc=wright.feng@cypress.com \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.