Skip to contentSkip to conversation

gcc-10 warning fixes

PATCH Applied Arnd Bergmann 14 participants 40 messages Apr 30, 2020 – Jun 10, 2020

Subsystems (22)
lore
Share & tools
lore
#0 lore raw

cover gcc-10 warning fixes

Here are a couple of fixes for warnings introduced with gcc-10. If you wish to reproduce these, you can find the compiler I used at [1].

If you like the fixes, please apply them directly into maintainer trees. I expect that we will also need them to be backported into stable kernels later.

I disabled -Wrestrict on gcc in my local test tree, but with the patches from this series and the ones I have already sent, I see no gcc-10 specific warnings in linux-next when doing many randconfig builds for arm/arm64/x86.

Arnd

Arnd Bergmann (15):
crypto - Avoid free() namespace collision
iwlwifi: mvm: fix gcc-10 zero-length-bounds warning
mwifiex: avoid -Wstringop-overflow warning
ath10k: fix gcc-10 zero-length-bounds warnings
bpf: avoid gcc-10 stringop-overflow warning
netfilter: conntrack: avoid gcc-10 zero-length-bounds warning
drop_monitor: work around gcc-10 stringop-overflow warning
usb: ehci: avoid gcc-10 zero-length-bounds warning
udf: avoid gcc-10 zero-length-bounds warnings
hpfs: avoid gcc-10 zero-length-bounds warning
omfs: avoid gcc-10 stringop-overflow warning
media: s5k5baf: avoid gcc-10 zero-length-bounds warning
scsi: sas: avoid gcc-10 zero-length-bounds warning
isci: avoid gcc-10 zero-length-bounds warning
nvme: avoid gcc-10 zero-length-bounds warning

 crypto/lrw.c                                  |  6 +--
 crypto/xts.c                                  |  6 +--
 drivers/media/i2c/s5k5baf.c                   |  4 +-
 drivers/net/wireless/ath/ath10k/htt.h         |  4 +-
 .../net/wireless/intel/iwlwifi/fw/api/tx.h    | 14 +++----
 .../net/wireless/marvell/mwifiex/sta_cmd.c    | 39 ++++++++-----------
 drivers/nvme/host/fc.c                        |  2 +-
 drivers/scsi/aic94xx/aic94xx_tmf.c            |  4 +-
 drivers/scsi/isci/task.h                      |  7 ++--
 drivers/scsi/libsas/sas_task.c                |  3 +-
 fs/hpfs/anode.c                               |  7 +++-
 fs/omfs/file.c                                | 12 +++---
 fs/omfs/omfs_fs.h                             |  2 +-
 fs/udf/ecma_167.h                             |  2 +-
 fs/udf/super.c                                |  2 +-
 include/linux/filter.h                        |  6 +--
 include/linux/usb/ehci_def.h                  | 12 ++++--
 include/net/netfilter/nf_conntrack.h          |  2 +-
 net/core/drop_monitor.c                       | 11 ++++--
 net/netfilter/nf_conntrack_core.c             |  4 +-
 20 files changed, 76 insertions(+), 73 deletions(-)

[1] https://mirrors.edge.kernel.org/pub/tools/crosstool/files/bin/x86_64/10.0.20200413/

Cc Herbert Xu <[email protected]>
Cc "David S. Miller" <[email protected]>
Cc Mauro Carvalho Chehab <[email protected]>
Cc Kalle Valo <[email protected]>
Cc Johannes Berg <[email protected]>
Cc Intel Linux Wireless <[email protected]>
Cc Amitkumar Karwar <[email protected]>
Cc James Smart <[email protected]>
Cc Jens Axboe <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc "James E.J. Bottomley" <[email protected]>
Cc "Martin K. Petersen" <[email protected]>
Cc Mikulas Patocka <[email protected]>
Cc Bob Copeland <[email protected]>
Cc Jan Kara <[email protected]>
Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
Cc Pablo Neira Ayuso <[email protected]>
Cc Florian Westphal <[email protected]>
Cc Jakub Kicinski <[email protected]>
Cc Neil Horman <[email protected]>
Signature · 7 lines
2.26.0


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#1 lore raw

PATCH 1/15 crypto - Avoid free() namespace collision

gcc-10 complains about using the name of a standard library function in the kernel, as we are not building with -ffreestanding:

crypto/xts.c:325:13: error: conflicting types for built-in function 'free'; expected 'void(void *)' [-Werror=builtin-declaration-mismatch]
325 | static void free(struct skcipher_instance *inst)

      |             ^~~~
crypto/lrw.c:290:13: error: conflicting types for built-in function 'free'; expected 'void(void *)' [-Werror=builtin-declaration-mismatch]
  290 | static void free(struct skcipher_instance *inst)
      |             ^~~~

crypto/lrw.c:27:1: note: 'free' is declared in header '<stdlib.h>'

The xts and lrw cipher implementations run into this because they do not use the conventional namespaced function names.

It might be better to rename all local functions in those files to help with things like 'ctags' and 'grep', but just renaming these two avoids the build issue. I picked the more verbose crypto_xts_free() and crypto_lrw_free() names for consistency with several other drivers that do use namespaced function names.

Fixes f1c131b45410 ("crypto: xts - Convert to skcipher")
Fixes 700cb3f5fe75 ("crypto: lrw - Convert to skcipher")
---
 crypto/lrw.c | 6 +++---
 crypto/xts.c | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)
crypto/lrw.c +3 -3
index 376d7ed3f1f8..5b07a7c09296 100644
@@ -287,7 +287,7 @@ static void exit_tfm(struct crypto_skcipher *tfm)
 	crypto_free_skcipher(ctx->child);
 }
 
-static void free(struct skcipher_instance *inst)
+static void crypto_lrw_free(struct skcipher_instance *inst)
 {
 	crypto_drop_skcipher(skcipher_instance_ctx(inst));
 	kfree(inst);
@@ -400,12 +400,12 @@ static int create(struct crypto_template *tmpl, struct rtattr **tb)
 	inst->alg.encrypt = encrypt;
 	inst->alg.decrypt = decrypt;
 
-	inst->free = free;
+	inst->free = crypto_lrw_free;
 
 	err = skcipher_register_instance(tmpl, inst);
 	if (err) {
 err_free_inst:
-		free(inst);
+		crypto_lrw_free(inst);
 	}
 	return err;
 }
crypto/xts.c +3 -3
index dbdd8af629e6..3565f3b863a6 100644
@@ -322,7 +322,7 @@ static void exit_tfm(struct crypto_skcipher *tfm)
 	crypto_free_cipher(ctx->tweak);
 }
 
-static void free(struct skcipher_instance *inst)
+static void crypto_xts_free(struct skcipher_instance *inst)
 {
 	crypto_drop_skcipher(skcipher_instance_ctx(inst));
 	kfree(inst);
@@ -434,12 +434,12 @@ static int create(struct crypto_template *tmpl, struct rtattr **tb)
 	inst->alg.encrypt = encrypt;
 	inst->alg.decrypt = decrypt;
 
-	inst->free = free;
+	inst->free = crypto_xts_free;
 
 	err = skcipher_register_instance(tmpl, inst);
 	if (err) {
 err_free_inst:
-		free(inst);
+		crypto_xts_free(inst);
 	}
 	return err;
 }
Signature · 1 line
2.26.0
maintainer
#2 lore raw

Re: [PATCH 01/15] crypto - Avoid free() namespace collision

Quoted 20 lines · Arnd Bergmann

gcc-10 complains about using the name of a standard library function in the kernel, as we are not building with -ffreestanding:

crypto/xts.c:325:13: error: conflicting types for built-in function 'free'; expected 'void(void *)' [-Werror=builtin-declaration-mismatch]
325 | static void free(struct skcipher_instance *inst)

      |             ^~~~
crypto/lrw.c:290:13: error: conflicting types for built-in function 'free'; expected 'void(void *)' [-Werror=builtin-declaration-mismatch]
  290 | static void free(struct skcipher_instance *inst)
      |             ^~~~

crypto/lrw.c:27:1: note: 'free' is declared in header '<stdlib.h>'

The xts and lrw cipher implementations run into this because they do not use the conventional namespaced function names.

It might be better to rename all local functions in those files to help with things like 'ctags' and 'grep', but just renaming these two avoids the build issue. I picked the more verbose crypto_xts_free() and crypto_lrw_free() names for consistency with several other drivers that do use namespaced function names.

Quoted 7 lines
Fixes f1c131b45410 ("crypto: xts - Convert to skcipher")
Fixes 700cb3f5fe75 ("crypto: lrw - Convert to skcipher")
---
 crypto/lrw.c | 6 +++---
 crypto/xts.c | 6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

Patch applied. Thanks.

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#3 lore raw

PATCH 2/15 iwlwifi: mvm: fix gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

gcc-10 complains when a zero-length array is accessed:

drivers/net/wireless/intel/iwlwifi/mvm/tx.c: In function 'iwl_mvm_rx_ba_notif':
drivers/net/wireless/intel/iwlwifi/mvm/tx.c:1929:17: warning: array subscript 9 is outside the bounds of an interior zero-length array 'struct iwl_mvm_compressed_ba_tfd[0]' [-Wzero-length-bounds]

 1929 |     &ba_res->tfd[i];
      |      ~~~~~~~~~~~^~~

In file included from drivers/net/wireless/intel/iwlwifi/mvm/../fw/api/tdls.h:68,
from drivers/net/wireless/intel/iwlwifi/mvm/fw-api.h:68, from drivers/net/wireless/intel/iwlwifi/mvm/sta.h:73,
from drivers/net/wireless/intel/iwlwifi/mvm/mvm.h:83,
from drivers/net/wireless/intel/iwlwifi/mvm/tx.c:72:
drivers/net/wireless/intel/iwlwifi/mvm/../fw/api/tx.h:769:35: note: while referencing 'tfd'
769 | struct iwl_mvm_compressed_ba_tfd tfd[0];
| ^~~

Change this structure to use a flexible-array member for 'tfd' instead, along with the various structures using an zero-length ieee80211_hdr array that do not show warnings today but might be affected by similar issues in the future.

Fixes 6f68cc367ab6 ("iwlwifi: api: annotate compressed BA notif array sizes")
Fixes c46e7724bfe9 ("iwlwifi: mvm: support new BA notification response")
---
 drivers/net/wireless/intel/iwlwifi/fw/api/tx.h | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
drivers/net/wireless/intel/iwlwifi/fw/api/tx.h +7 -7
index f1d1fe96fecc..82d59b5a5f8c 100644
@@ -293,7 +293,7 @@ struct iwl_tx_cmd {
 	__le16 pm_frame_timeout;
 	__le16 reserved4;
 	u8 payload[0];
-	struct ieee80211_hdr hdr[0];
+	struct ieee80211_hdr hdr[];
 } __packed; /* TX_CMD_API_S_VER_6 */
 
 struct iwl_dram_sec_info {
@@ -319,7 +319,7 @@ struct iwl_tx_cmd_gen2 {
 	__le32 flags;
 	struct iwl_dram_sec_info dram_info;
 	__le32 rate_n_flags;
-	struct ieee80211_hdr hdr[0];
+	struct ieee80211_hdr hdr[];
 } __packed; /* TX_CMD_API_S_VER_7 */
 
 /**
@@ -342,7 +342,7 @@ struct iwl_tx_cmd_gen3 {
 	struct iwl_dram_sec_info dram_info;
 	__le32 rate_n_flags;
 	__le64 ttl;
-	struct ieee80211_hdr hdr[0];
+	struct ieee80211_hdr hdr[];
 } __packed; /* TX_CMD_API_S_VER_8 */
 
 /*
@@ -766,8 +766,8 @@ struct iwl_mvm_compressed_ba_notif {
 	__le32 tx_rate;
 	__le16 tfd_cnt;
 	__le16 ra_tid_cnt;
-	struct iwl_mvm_compressed_ba_tfd tfd[0];
 	struct iwl_mvm_compressed_ba_ratid ra_tid[0];
+	struct iwl_mvm_compressed_ba_tfd tfd[];
 } __packed; /* COMPRESSED_BA_RES_API_S_VER_4 */
 
 /**
@@ -784,7 +784,7 @@ struct iwl_mac_beacon_cmd_v6 {
 	__le32 template_id;
 	__le32 tim_idx;
 	__le32 tim_size;
-	struct ieee80211_hdr frame[0];
+	struct ieee80211_hdr frame[];
 } __packed; /* BEACON_TEMPLATE_CMD_API_S_VER_6 */
 
 /**
@@ -805,7 +805,7 @@ struct iwl_mac_beacon_cmd_v7 {
 	__le32 tim_size;
 	__le32 ecsa_offset;
 	__le32 csa_offset;
-	struct ieee80211_hdr frame[0];
+	struct ieee80211_hdr frame[];
 } __packed; /* BEACON_TEMPLATE_CMD_API_S_VER_7 */
 
 enum iwl_mac_beacon_flags {
@@ -840,7 +840,7 @@ struct iwl_mac_beacon_cmd {
 	__le32 tim_size;
 	__le32 ecsa_offset;
 	__le32 csa_offset;
-	struct ieee80211_hdr frame[0];
+	struct ieee80211_hdr frame[];
 } __packed; /* BEACON_TEMPLATE_CMD_API_S_VER_10 */
 
 struct iwl_beacon_notif {
Signature · 1 line
2.26.0
#4 lore raw

Re: [PATCH 02/15] iwlwifi: mvm: fix gcc-10 zero-length-bounds warning

Quoted 15 lines · Arnd Bergmann

gcc-10 complains when a zero-length array is accessed:

drivers/net/wireless/intel/iwlwifi/mvm/tx.c: In function 'iwl_mvm_rx_ba_notif':
drivers/net/wireless/intel/iwlwifi/mvm/tx.c:1929:17: warning: array subscript 9 is outside the bounds of an interior zero-length array 'struct iwl_mvm_compressed_ba_tfd[0]' [-Wzero-length-bounds]

 1929 |     &ba_res->tfd[i];
      |      ~~~~~~~~~~~^~~

In file included from drivers/net/wireless/intel/iwlwifi/mvm/../fw/api/tdls.h:68,
from drivers/net/wireless/intel/iwlwifi/mvm/fw-api.h:68, from drivers/net/wireless/intel/iwlwifi/mvm/sta.h:73,
from drivers/net/wireless/intel/iwlwifi/mvm/mvm.h:83,
from drivers/net/wireless/intel/iwlwifi/mvm/tx.c:72:
drivers/net/wireless/intel/iwlwifi/mvm/../fw/api/tx.h:769:35: note: while referencing 'tfd'
769 | struct iwl_mvm_compressed_ba_tfd tfd[0];
| ^~~

Quoted 8 lines

Change this structure to use a flexible-array member for 'tfd' instead, along with the various structures using an zero-length ieee80211_hdr array that do not show warnings today but might be affected by similar issues in the future.

Fixes 6f68cc367ab6 ("iwlwifi: api: annotate compressed BA notif array sizes")
Fixes c46e7724bfe9 ("iwlwifi: mvm: support new BA notification response")

Patch applied to iwlwifi-next.git, thanks.

9cec1d547cb7 iwlwifi: mvm: fix gcc-10 zero-length-bounds warning

#5 lore raw

PATCH 3/15 mwifiex: avoid -Wstringop-overflow warning

In reply to Arnd Bergmann

gcc-10 reports a warning for mwifiex_cmd_802_11_key_material_v1:

drivers/net/wireless/marvell/mwifiex/sta_cmd.c: In function 'mwifiex_cmd_802_11_key_material_v1':
cc1: warning: writing 16 bytes into a region of size 0 [-Wstringop-overflow=]
In file included from drivers/net/wireless/marvell/mwifiex/sta_cmd.c:23:
drivers/net/wireless/marvell/mwifiex/fw.h:993:9: note: at offset 0 to object 'action' with size 2 declared here

  993 |  __le16 action;
      |         ^~~~~~

As the warning makes no sense, I reported it as a bug for gcc. In the meantime using a temporary pointer for the key data makes the code easier to read and stops the warning.

Fixes 5e6e3a92b9a4 ("wireless: mwifiex: initial commit for Marvell mwifiex driver")
---
 .../net/wireless/marvell/mwifiex/sta_cmd.c    | 39 ++++++++-----------
 1 file changed, 16 insertions(+), 23 deletions(-)
drivers/net/wireless/marvell/mwifiex/sta_cmd.c +16 -23
index 0bd93f26bd7f..8bd355d7974e 100644
@@ -853,43 +853,36 @@ mwifiex_cmd_802_11_key_material_v1(struct mwifiex_private *priv,
 		memset(&key_material->key_param_set, 0,
 		       sizeof(struct mwifiex_ie_type_key_param_set));
 	if (enc_key->is_wapi_key) {
+		struct mwifiex_ie_type_key_param_set *set;
+
 		mwifiex_dbg(priv->adapter, INFO, "info: Set WAPI Key\n");
-		key_material->key_param_set.key_type_id =
-						cpu_to_le16(KEY_TYPE_ID_WAPI);
+		set = &key_material->key_param_set;
+		set->key_type_id = cpu_to_le16(KEY_TYPE_ID_WAPI);
 		if (cmd_oid == KEY_INFO_ENABLED)
-			key_material->key_param_set.key_info =
-						cpu_to_le16(KEY_ENABLED);
+			set->key_info = cpu_to_le16(KEY_ENABLED);
 		else
-			key_material->key_param_set.key_info =
-						cpu_to_le16(!KEY_ENABLED);
+			set->key_info = cpu_to_le16(!KEY_ENABLED);
 
-		key_material->key_param_set.key[0] = enc_key->key_index;
+		set->key[0] = enc_key->key_index;
 		if (!priv->sec_info.wapi_key_on)
-			key_material->key_param_set.key[1] = 1;
+			set->key[1] = 1;
 		else
 			/* set 0 when re-key */
-			key_material->key_param_set.key[1] = 0;
+			set->key[1] = 0;
 
 		if (!is_broadcast_ether_addr(enc_key->mac_addr)) {
 			/* WAPI pairwise key: unicast */
-			key_material->key_param_set.key_info |=
-				cpu_to_le16(KEY_UNICAST);
+			set->key_info |= cpu_to_le16(KEY_UNICAST);
 		} else {	/* WAPI group key: multicast */
-			key_material->key_param_set.key_info |=
-				cpu_to_le16(KEY_MCAST);
+			set->key_info |= cpu_to_le16(KEY_MCAST);
 			priv->sec_info.wapi_key_on = true;
 		}
 
-		key_material->key_param_set.type =
-					cpu_to_le16(TLV_TYPE_KEY_MATERIAL);
-		key_material->key_param_set.key_len =
-						cpu_to_le16(WAPI_KEY_LEN);
-		memcpy(&key_material->key_param_set.key[2],
-		       enc_key->key_material, enc_key->key_len);
-		memcpy(&key_material->key_param_set.key[2 + enc_key->key_len],
-		       enc_key->pn, PN_LEN);
-		key_material->key_param_set.length =
-			cpu_to_le16(WAPI_KEY_LEN + KEYPARAMSET_FIXED_LEN);
+		set->type = cpu_to_le16(TLV_TYPE_KEY_MATERIAL);
+		set->key_len = cpu_to_le16(WAPI_KEY_LEN);
+		memcpy(&set->key[2], enc_key->key_material, enc_key->key_len);
+		memcpy(&set->key[2 + enc_key->key_len], enc_key->pn, PN_LEN);
+		set->length = cpu_to_le16(WAPI_KEY_LEN + KEYPARAMSET_FIXED_LEN);
 
 		key_param_len = (WAPI_KEY_LEN + KEYPARAMSET_FIXED_LEN) +
 				 sizeof(struct mwifiex_ie_types_header);
Signature · 1 line
2.26.0
#6 lore raw

Re: [PATCH 03/15] mwifiex: avoid -Wstringop-overflow warning

Quoted 9 lines · Arnd Bergmann

gcc-10 reports a warning for mwifiex_cmd_802_11_key_material_v1:

drivers/net/wireless/marvell/mwifiex/sta_cmd.c: In function 'mwifiex_cmd_802_11_key_material_v1':
cc1: warning: writing 16 bytes into a region of size 0 [-Wstringop-overflow=]
In file included from drivers/net/wireless/marvell/mwifiex/sta_cmd.c:23:
drivers/net/wireless/marvell/mwifiex/fw.h:993:9: note: at offset 0 to object 'action' with size 2 declared here

  993 |  __le16 action;
      |         ^~~~~~
Quoted 7 lines

As the warning makes no sense, I reported it as a bug for gcc. In the meantime using a temporary pointer for the key data makes the code easier to read and stops the warning.

Fixes 5e6e3a92b9a4 ("wireless: mwifiex: initial commit for Marvell mwifiex driver")

Patch applied to wireless-drivers-next.git, thanks.

08afb432c996 mwifiex: avoid -Wstringop-overflow warning

Signature · 3 lines
https://patchwork.kernel.org/patch/11521661/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
#7 lore raw

PATCH 4/15 ath10k: fix gcc-10 zero-length-bounds warnings

In reply to Arnd Bergmann

gcc-10 started warning about out-of-bounds access for zero-length arrays:

In file included from drivers/net/wireless/ath/ath10k/core.h:18,
from drivers/net/wireless/ath/ath10k/htt_rx.c:8:
drivers/net/wireless/ath/ath10k/htt_rx.c: In function 'ath10k_htt_rx_tx_fetch_ind':
drivers/net/wireless/ath/ath10k/htt.h:1683:17: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct htt_tx_fetch_record[0]' [-Wzero-length-bounds]

 1683 |  return (void *)&ind->records[le16_to_cpu(ind->num_records)];
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

drivers/net/wireless/ath/ath10k/htt.h:1676:29: note: while referencing 'records'
1676 | struct htt_tx_fetch_record records[0];
| ^~~~~~~

The structure was already converted to have a flexible-array member in the past, but there are two zero-length members in the end and only one of them can be a flexible-array member.

Swap the two around to avoid the warning, as 'resp_ids' is not accessed in a way that causes a warning.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 22e6b3bc5d96 ("ath10k: add new htt definitions")
---
 drivers/net/wireless/ath/ath10k/htt.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/wireless/ath/ath10k/htt.h +2 -2
index e7096a73c6ca..7621f0a3dc77 100644
@@ -1673,8 +1673,8 @@ struct htt_tx_fetch_ind {
 	__le32 token;
 	__le16 num_resp_ids;
 	__le16 num_records;
-	struct htt_tx_fetch_record records[0];
-	__le32 resp_ids[]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	__le32 resp_ids[0]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	struct htt_tx_fetch_record records[];
 } __packed;
 
 static inline void *
Signature · 7 lines
2.26.0


_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#8 lore raw

Re: [PATCH 04/15] ath10k: fix gcc-10 zero-length-bounds warnings

Hi Arnd,

Quoted 33 lines · Arnd Bergmann

gcc-10 started warning about out-of-bounds access for zero-length arrays:

In file included from drivers/net/wireless/ath/ath10k/core.h:18,
from drivers/net/wireless/ath/ath10k/htt_rx.c:8:
drivers/net/wireless/ath/ath10k/htt_rx.c: In function 'ath10k_htt_rx_tx_fetch_ind':
drivers/net/wireless/ath/ath10k/htt.h:1683:17: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct htt_tx_fetch_record[0]' [-Wzero-length-bounds]

 1683 |  return (void *)&ind->records[le16_to_cpu(ind->num_records)];
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

drivers/net/wireless/ath/ath10k/htt.h:1676:29: note: while referencing 'records'
1676 | struct htt_tx_fetch_record records[0];
| ^~~~~~~

The structure was already converted to have a flexible-array member in the past, but there are two zero-length members in the end and only one of them can be a flexible-array member.

Swap the two around to avoid the warning, as 'resp_ids' is not accessed in a way that causes a warning.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 22e6b3bc5d96 ("ath10k: add new htt definitions")
---
 drivers/net/wireless/ath/ath10k/htt.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/wireless/ath/ath10k/htt.h +0 -0
index e7096a73c6ca..7621f0a3dc77 100644

drivers/net/wireless/ath/ath10k/htt.h struct htt_tx_fetch_ind { · line 1680

Quoted 10 lines
drivers/net/wireless/ath/ath10k/htt.h +2 -2
@@ -1673,8 +1673,8 @@ struct htt_tx_fetch_ind {
 	__le32 token;
 	__le16 num_resp_ids;
 	__le16 num_records;
-	struct htt_tx_fetch_record records[0];
-	__le32 resp_ids[]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	__le32 resp_ids[0]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	struct htt_tx_fetch_record records[];
 } __packed;
 
 static inline void *

The treewide patch is an experimental change and, as this change only applies to my -next tree, I will carry this patch in it, so other people don't have to worry about this at all.

Thank you
--
Gustavo

#9 lore raw

Re: [PATCH 04/15] ath10k: fix gcc-10 zero-length-bounds warnings

Quoted 6 lines · Gustavo A. R. Silva
Quoted 2 lines · Arnd Bergmann

gcc-10 started warning about out-of-bounds access for zero-length arrays:

The treewide patch is an experimental change and, as this change only applies to my -next tree, I will carry this patch in it, so other people don't have to worry about this at all.

Ok, thanks!

Arnd

_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k

#10 lore raw

Re: [PATCH 04/15] ath10k: fix gcc-10 zero-length-bounds warnings

In reply to Gustavo A. R. Silva
Quoted 44 lines · Gustavo A. R. Silva

Hi Arnd,

Quoted 41 lines · Arnd Bergmann

gcc-10 started warning about out-of-bounds access for zero-length arrays:

In file included from drivers/net/wireless/ath/ath10k/core.h:18,
from drivers/net/wireless/ath/ath10k/htt_rx.c:8:
drivers/net/wireless/ath/ath10k/htt_rx.c: In function 'ath10k_htt_rx_tx_fetch_ind':
drivers/net/wireless/ath/ath10k/htt.h:1683:17: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct htt_tx_fetch_record[0]' [-Wzero-length-bounds]

 1683 |  return (void *)&ind->records[le16_to_cpu(ind->num_records)];
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

drivers/net/wireless/ath/ath10k/htt.h:1676:29: note: while referencing 'records'
1676 | struct htt_tx_fetch_record records[0];
| ^~~~~~~

The structure was already converted to have a flexible-array member in the past, but there are two zero-length members in the end and only one of them can be a flexible-array member.

Swap the two around to avoid the warning, as 'resp_ids' is not accessed in a way that causes a warning.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 22e6b3bc5d96 ("ath10k: add new htt definitions")
---
 drivers/net/wireless/ath/ath10k/htt.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/wireless/ath/ath10k/htt.h +2 -2
index e7096a73c6ca..7621f0a3dc77 100644
@@ -1673,8 +1673,8 @@ struct htt_tx_fetch_ind {
 	__le32 token;
 	__le16 num_resp_ids;
 	__le16 num_records;
-	struct htt_tx_fetch_record records[0];
-	__le32 resp_ids[]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	__le32 resp_ids[0]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	struct htt_tx_fetch_record records[];
 } __packed;
 
Quoted 6 lines

drivers/net/wireless/ath/ath10k/htt.h struct htt_tx_fetch_ind { · line 1680

Quoted 2 lines
drivers/net/wireless/ath/ath10k/htt.h +0 -0
 static inline void *

The treewide patch is an experimental change and, as this change only applies to my -next tree, I will carry this patch in it, so other people don't have to worry about this at all.

Gustavo, why do you have ath10k patches in your tree? I prefer that ath10k patches go through my ath.git tree so that they are reviewed and tested.

Signature · 6 lines
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#11 lore raw

Re: [PATCH 04/15] ath10k: fix gcc-10 zero-length-bounds warnings

Quoted 53 lines · Kalle Valo
Quoted 44 lines · Gustavo A. R. Silva

Hi Arnd,

Quoted 41 lines · Arnd Bergmann

gcc-10 started warning about out-of-bounds access for zero-length arrays:

In file included from drivers/net/wireless/ath/ath10k/core.h:18,
from drivers/net/wireless/ath/ath10k/htt_rx.c:8:
drivers/net/wireless/ath/ath10k/htt_rx.c: In function 'ath10k_htt_rx_tx_fetch_ind':
drivers/net/wireless/ath/ath10k/htt.h:1683:17: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct htt_tx_fetch_record[0]' [-Wzero-length-bounds]

 1683 |  return (void *)&ind->records[le16_to_cpu(ind->num_records)];
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

drivers/net/wireless/ath/ath10k/htt.h:1676:29: note: while referencing 'records'
1676 | struct htt_tx_fetch_record records[0];
| ^~~~~~~

The structure was already converted to have a flexible-array member in the past, but there are two zero-length members in the end and only one of them can be a flexible-array member.

Swap the two around to avoid the warning, as 'resp_ids' is not accessed in a way that causes a warning.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 22e6b3bc5d96 ("ath10k: add new htt definitions")
---
 drivers/net/wireless/ath/ath10k/htt.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/wireless/ath/ath10k/htt.h +2 -2
index e7096a73c6ca..7621f0a3dc77 100644
@@ -1673,8 +1673,8 @@ struct htt_tx_fetch_ind {
 	__le32 token;
 	__le16 num_resp_ids;
 	__le16 num_records;
-	struct htt_tx_fetch_record records[0];
-	__le32 resp_ids[]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	__le32 resp_ids[0]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	struct htt_tx_fetch_record records[];
 } __packed;
 
Quoted 6 lines

drivers/net/wireless/ath/ath10k/htt.h struct htt_tx_fetch_ind { · line 1680

Quoted 2 lines
drivers/net/wireless/ath/ath10k/htt.h +0 -0
 static inline void *

The treewide patch is an experimental change and, as this change only applies to my -next tree, I will carry this patch in it, so other people don't have to worry about this at all.

Quoted 4 lines

Gustavo, why do you have ath10k patches in your tree? I prefer that ath10k patches go through my ath.git tree so that they are reviewed and tested.

I just wanted to test out a mechanical change. I will remove it from my tree now and will send a patch to you so you can apply it to your ath.git tree.

Thanks!
--
Gustavo

#12 lore raw

Re: [PATCH 04/15] ath10k: fix gcc-10 zero-length-bounds warnings

Quoted 54 lines · Gustavo A. R. Silva
Quoted 53 lines · Kalle Valo
Quoted 50 lines · Gustavo A. R. Silva

Hi Arnd,

Quoted 33 lines · Arnd Bergmann

gcc-10 started warning about out-of-bounds access for zero-length arrays:

In file included from drivers/net/wireless/ath/ath10k/core.h:18,
from drivers/net/wireless/ath/ath10k/htt_rx.c:8:
drivers/net/wireless/ath/ath10k/htt_rx.c: In function 'ath10k_htt_rx_tx_fetch_ind':
drivers/net/wireless/ath/ath10k/htt.h:1683:17: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct htt_tx_fetch_record[0]' [-Wzero-length-bounds]

 1683 |  return (void *)&ind->records[le16_to_cpu(ind->num_records)];
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

drivers/net/wireless/ath/ath10k/htt.h:1676:29: note: while referencing 'records'
1676 | struct htt_tx_fetch_record records[0];
| ^~~~~~~

The structure was already converted to have a flexible-array member in the past, but there are two zero-length members in the end and only one of them can be a flexible-array member.

Swap the two around to avoid the warning, as 'resp_ids' is not accessed in a way that causes a warning.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 22e6b3bc5d96 ("ath10k: add new htt definitions")
---
 drivers/net/wireless/ath/ath10k/htt.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/wireless/ath/ath10k/htt.h +0 -0
index e7096a73c6ca..7621f0a3dc77 100644

drivers/net/wireless/ath/ath10k/htt.h struct htt_tx_fetch_ind { · line 1680

Quoted 10 lines
drivers/net/wireless/ath/ath10k/htt.h +2 -2
@@ -1673,8 +1673,8 @@ struct htt_tx_fetch_ind {
 	__le32 token;
 	__le16 num_resp_ids;
 	__le16 num_records;
-	struct htt_tx_fetch_record records[0];
-	__le32 resp_ids[]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	__le32 resp_ids[0]; /* ath10k_htt_get_tx_fetch_ind_resp_ids() */
+	struct htt_tx_fetch_record records[];
 } __packed;
 
 static inline void *

The treewide patch is an experimental change and, as this change only applies to my -next tree, I will carry this patch in it, so other people don't have to worry about this at all.

Quoted 7 lines
Quoted 4 lines

Gustavo, why do you have ath10k patches in your tree? I prefer that ath10k patches go through my ath.git tree so that they are reviewed and tested.

I just wanted to test out a mechanical change. I will remove it from my tree now and will send a patch to you so you can apply it to your ath.git tree.

Great, thanks.

Signature · 6 lines
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

_______________________________________________
ath10k mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/ath10k
#13 lore raw

PATCH 5/15 bpf: avoid gcc-10 stringop-overflow warning

In reply to Arnd Bergmann

gcc-10 warns about accesses to zero-length arrays:

kernel/bpf/core.c: In function 'bpf_patch_insn_single':
cc1: warning: writing 8 bytes into a region of size 0 [-Wstringop-overflow=] In file included from kernel/bpf/core.c:21:
include/linux/filter.h:550:20: note: at offset 0 to object 'insnsi' with size 0 declared here

  550 |   struct bpf_insn  insnsi[0];
      |                    ^~~~~~

In this case, we really want to have two flexible-array members, but that is not possible. Removing the union to make insnsi a flexible-array member while leaving insns as a zero-length array fixes the warning, as nothing writes to the other one in that way.

This trick only works on linux-3.18 or higher, as older versions had additional members in the union.

Fixes 60a3b2253c41 ("net: bpf: make eBPF interpreter images read-only")
---
 include/linux/filter.h | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
include/linux/filter.h +2 -4
index af37318bb1c5..73d06a39e2d6 100644
@@ -545,10 +545,8 @@ struct bpf_prog {
 	unsigned int		(*bpf_func)(const void *ctx,
 					    const struct bpf_insn *insn);
 	/* Instructions for interpreter */
-	union {
-		struct sock_filter	insns[0];
-		struct bpf_insn		insnsi[0];
-	};
+	struct sock_filter	insns[0];
+	struct bpf_insn		insnsi[];
 };
 
 struct sk_filter {
Signature · 1 line
2.26.0
maintainer Applied
#14 lore raw

Re: [PATCH 05/15] bpf: avoid gcc-10 stringop-overflow warning

Quoted 14 lines · Arnd Bergmann

gcc-10 warns about accesses to zero-length arrays:

kernel/bpf/core.c: In function 'bpf_patch_insn_single':
cc1: warning: writing 8 bytes into a region of size 0 [-Wstringop-overflow=] In file included from kernel/bpf/core.c:21:
include/linux/filter.h:550:20: note: at offset 0 to object 'insnsi' with size 0 declared here

   550 |   struct bpf_insn  insnsi[0];
       |                    ^~~~~~

In this case, we really want to have two flexible-array members, but that is not possible. Removing the union to make insnsi a flexible-array member while leaving insns as a zero-length array fixes the warning, as nothing writes to the other one in that way.

Quoted 5 lines

This trick only works on linux-3.18 or higher, as older versions had additional members in the union.

Fixes 60a3b2253c41 ("net: bpf: make eBPF interpreter images read-only")

Not pretty but looks okay to me, both have the same offset afterwards in the pahole dump as well.

struct bpf_prog {
[...]

        unsigned int               (*bpf_func)(const void  *, const struct bpf_insn  *); /*    48     8 */
        struct sock_filter         insns[0];             /*    56     0 */
        struct bpf_insn            insnsi[];             /*    56     0 */
        /* size: 56, cachelines: 1, members: 21 */
        /* sum members: 50, holes: 1, sum holes: 4 */
        /* sum bitfield members: 10 bits, bit holes: 1, sum bit holes: 6 bits */
        /* last cacheline: 56 bytes */
};

Applied to bpf-next, thanks!

#15 lore raw

PATCH 6/15 netfilter: conntrack: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

gcc-10 warns around a suspicious access to an empty struct member:

net/netfilter/nf_conntrack_core.c: In function '__nf_conntrack_alloc':
net/netfilter/nf_conntrack_core.c:1522:9: warning: array subscript 0 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]
1522 | memset(&ct->__nfct_init_offset[0], 0,

      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from net/netfilter/nf_conntrack_core.c:37:
include/net/netfilter/nf_conntrack.h:90:5: note: while referencing '__nfct_init_offset'
   90 |  u8 __nfct_init_offset[0];
      |     ^~~~~~~~~~~~~~~~~~

The code is correct but a bit unusual. Rework it slightly in a way that does not trigger the warning, using an empty struct instead of an empty array. There are probably more elegant ways to do this, but this is the smallest change.

Fixes c41884ce0562 ("netfilter: conntrack: avoid zeroing timer")
---
 include/net/netfilter/nf_conntrack.h | 2 +-
 net/netfilter/nf_conntrack_core.c    | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)
include/net/netfilter/nf_conntrack.h +1 -1
index 9f551f3b69c6..90690e37a56f 100644
@@ -87,7 +87,7 @@ struct nf_conn {
 	struct hlist_node	nat_bysource;
 #endif
 	/* all members below initialized via memset */
-	u8 __nfct_init_offset[0];
+	struct { } __nfct_init_offset;
 
 	/* If we were expected by an expectation, this will be it */
 	struct nf_conn *master;
net/netfilter/nf_conntrack_core.c +2 -2
index c4582eb71766..0173398f4ced 100644
@@ -1519,9 +1519,9 @@ __nf_conntrack_alloc(struct net *net,
 	ct->status = 0;
 	ct->timeout = 0;
 	write_pnet(&ct->ct_net, net);
-	memset(&ct->__nfct_init_offset[0], 0,
+	memset(&ct->__nfct_init_offset, 0,
 	       offsetof(struct nf_conn, proto) -
-	       offsetof(struct nf_conn, __nfct_init_offset[0]));
+	       offsetof(struct nf_conn, __nfct_init_offset));
 
 	nf_ct_zone_add(ct, zone);
 
Signature · 1 line
2.26.0
maintainer Applied
#16 lore raw

Re: [PATCH 06/15] netfilter: conntrack: avoid gcc-10 zero-length-bounds warning

Quoted 11 lines · Arnd Bergmann

gcc-10 warns around a suspicious access to an empty struct member:

net/netfilter/nf_conntrack_core.c: In function '__nf_conntrack_alloc':
net/netfilter/nf_conntrack_core.c:1522:9: warning: array subscript 0 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]
1522 | memset(&ct->__nfct_init_offset[0], 0,

      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from net/netfilter/nf_conntrack_core.c:37:
include/net/netfilter/nf_conntrack.h:90:5: note: while referencing '__nfct_init_offset'
   90 |  u8 __nfct_init_offset[0];
      |     ^~~~~~~~~~~~~~~~~~
Quoted 4 lines

The code is correct but a bit unusual. Rework it slightly in a way that does not trigger the warning, using an empty struct instead of an empty array. There are probably more elegant ways to do this, but this is the smallest change.

Applied, thanks.

#17 lore raw

PATCH 7/15 drop_monitor: work around gcc-10 stringop-overflow warning

In reply to Arnd Bergmann

The current gcc-10 snapshot produces a false-positive warning:

net/core/drop_monitor.c: In function 'trace_drop_common.constprop':
cc1: error: writing 8 bytes into a region of size 0 [-Werror=stringop-overflow=]
In file included from net/core/drop_monitor.c:23:
include/uapi/linux/net_dropmon.h:36:8: note: at offset 0 to object 'entries' with size 4 declared here

   36 |  __u32 entries;
      |        ^~~~~~~

I reported this in the gcc bugzilla, but in case it does not get fixed in the release, work around it by using a temporary variable.

Fixes 9a8afc8d3962 ("Network Drop Monitor: Adding drop monitor implementation & Netlink protocol")
---
 net/core/drop_monitor.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
net/core/drop_monitor.c +7 -4
index 8e33cec9fc4e..2ee7bc4c9e03 100644
@@ -213,6 +213,7 @@ static void sched_send_work(struct timer_list *t)
 static void trace_drop_common(struct sk_buff *skb, void *location)
 {
 	struct net_dm_alert_msg *msg;
+	struct net_dm_drop_point *point;
 	struct nlmsghdr *nlh;
 	struct nlattr *nla;
 	int i;
@@ -231,11 +232,13 @@ static void trace_drop_common(struct sk_buff *skb, void *location)
 	nlh = (struct nlmsghdr *)dskb->data;
 	nla = genlmsg_data(nlmsg_data(nlh));
 	msg = nla_data(nla);
+	point = msg->points;
 	for (i = 0; i < msg->entries; i++) {
-		if (!memcmp(&location, msg->points[i].pc, sizeof(void *))) {
-			msg->points[i].count++;
+		if (!memcmp(&location, &point->pc, sizeof(void *))) {
+			point->count++;
 			goto out;
 		}
+		point++;
 	}
 	if (msg->entries == dm_hit_limit)
 		goto out;
@@ -244,8 +247,8 @@ static void trace_drop_common(struct sk_buff *skb, void *location)
 	 */
 	__nla_reserve_nohdr(dskb, sizeof(struct net_dm_drop_point));
 	nla->nla_len += NLA_ALIGN(sizeof(struct net_dm_drop_point));
-	memcpy(msg->points[msg->entries].pc, &location, sizeof(void *));
-	msg->points[msg->entries].count = 1;
+	memcpy(point->pc, &location, sizeof(void *));
+	point->count = 1;
 	msg->entries++;
 
 	if (!timer_pending(&data->send_timer)) {
Signature · 1 line
2.26.0
maintainer Acked-by on v1 7/15
#18 lore raw

Re: [PATCH 07/15] drop_monitor: work around gcc-10 stringop-overflow warning

Quoted 48 lines · Arnd Bergmann

The current gcc-10 snapshot produces a false-positive warning:

net/core/drop_monitor.c: In function 'trace_drop_common.constprop':
cc1: error: writing 8 bytes into a region of size 0 [-Werror=stringop-overflow=]
In file included from net/core/drop_monitor.c:23:
include/uapi/linux/net_dropmon.h:36:8: note: at offset 0 to object 'entries' with size 4 declared here

   36 |  __u32 entries;
      |        ^~~~~~~

I reported this in the gcc bugzilla, but in case it does not get fixed in the release, work around it by using a temporary variable.

Fixes 9a8afc8d3962 ("Network Drop Monitor: Adding drop monitor implementation & Netlink protocol")
---
 net/core/drop_monitor.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
net/core/drop_monitor.c +5 -2
index 8e33cec9fc4e..2ee7bc4c9e03 100644
@@ -213,6 +213,7 @@ static void sched_send_work(struct timer_list *t)
 static void trace_drop_common(struct sk_buff *skb, void *location)
 {
 	struct net_dm_alert_msg *msg;
+	struct net_dm_drop_point *point;
 	struct nlmsghdr *nlh;
 	struct nlattr *nla;
 	int i;
@@ -231,11 +232,13 @@ static void trace_drop_common(struct sk_buff *skb, void *location)
 	nlh = (struct nlmsghdr *)dskb->data;
 	nla = genlmsg_data(nlmsg_data(nlh));
 	msg = nla_data(nla);
+	point = msg->points;
 	for (i = 0; i < msg->entries; i++) {
-		if (!memcmp(&location, msg->points[i].pc, sizeof(void *))) {
-			msg->points[i].count++;
+		if (!memcmp(&location, &point->pc, sizeof(void *))) {
+			point->count++;
 			goto out;
 		}
+		point++;
 	}
 	if (msg->entries == dm_hit_limit)
 		goto out;

net/core/drop_monitor.c static void trace_drop_common(struct sk_buff *skb, void *location) · line 254

Quoted 10 lines
net/core/drop_monitor.c +2 -2
@@ -244,8 +247,8 @@ static void trace_drop_common(struct sk_buff *skb, void *location)
 	 */
 	__nla_reserve_nohdr(dskb, sizeof(struct net_dm_drop_point));
 	nla->nla_len += NLA_ALIGN(sizeof(struct net_dm_drop_point));
-	memcpy(msg->points[msg->entries].pc, &location, sizeof(void *));
-	msg->points[msg->entries].count = 1;
+	memcpy(point->pc, &location, sizeof(void *));
+	point->count = 1;
 	msg->entries++;
 
 	if (!timer_pending(&data->send_timer)) {
#19 lore raw

PATCH 8/15 usb: ehci: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

Building ehci drivers with gcc-10 results in a number of warnings like when an zero-length array is accessed:

drivers/usb/host/ehci-hub.c: In function 'ehci_bus_suspend':
drivers/usb/host/ehci-hub.c:320:30: error: array subscript 14 is outside the bounds of an interior zero-length array 'u32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]

  320 |    u32 __iomem *hostpc_reg = &ehci->regs->hostpc[port];
      |                              ^~~~~~~~~~~~~~~~~~~~~~~~~
In file included from drivers/usb/host/ehci.h:273,
                 from drivers/usb/host/ehci-hcd.c:96:
include/linux/usb/ehci_def.h:186:7: note: while referencing 'hostpc'
  186 |  u32  hostpc[0]; /* HOSTPC extension */
      |       ^~~~~~
In file included from drivers/usb/host/ehci-hcd.c:305:
drivers/usb/host/ehci-hub.c: In function 'ehci_hub_control':
drivers/usb/host/ehci-hub.c:892:15: error: array subscript 256 is outside the bounds of an interior zero-length array 'u32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]
  892 |  hostpc_reg = &ehci->regs->hostpc[temp];
      |               ^~~~~~~~~~~~~~~~~~~~~~~~~
In file included from drivers/usb/host/ehci.h:273,
                 from drivers/usb/host/ehci-hcd.c:96:
include/linux/usb/ehci_def.h:186:7: note: while referencing 'hostpc'
  186 |  u32  hostpc[0]; /* HOSTPC extension */
      |       ^~~~~~

All these fields are colocated with reserved fields that I guess refer to the correct field length.

Change the two struct definition to use an unnamed union to define both of these fields at the same location as the corresponding reserved fields.

---
 include/linux/usb/ehci_def.h | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
include/linux/usb/ehci_def.h +8 -4
index 78e006355557..8777d8e56ef2 100644
@@ -127,7 +127,8 @@ struct ehci_regs {
 #define FLAG_CF		(1<<0)		/* true: we'll support "high speed" */
 
 	/* PORTSC: offset 0x44 */
-	u32		port_status[0];	/* up to N_PORTS */
+	union {
+		u32		port_status[9];	/* up to N_PORTS */
 /* EHCI 1.1 addendum */
 #define PORTSC_SUSPEND_STS_ACK 0
 #define PORTSC_SUSPEND_STS_NYET 1
@@ -165,7 +166,8 @@ struct ehci_regs {
 #define PORT_CONNECT	(1<<0)		/* device connected */
 #define PORT_RWC_BITS   (PORT_CSC | PORT_PEC | PORT_OCC)
 
-	u32		reserved3[9];
+		u32		reserved3[9];
+	};
 
 	/* USBMODE: offset 0x68 */
 	u32		usbmode;	/* USB Device mode */
@@ -181,11 +183,13 @@ struct ehci_regs {
  * PORTSCx
  */
 	/* HOSTPC: offset 0x84 */
-	u32		hostpc[0];	/* HOSTPC extension */
+	union {
+		u32		hostpc[17];	/* HOSTPC extension */
 #define HOSTPC_PHCD	(1<<22)		/* Phy clock disable */
 #define HOSTPC_PSPD	(3<<25)		/* Port speed detection */
 
-	u32		reserved5[17];
+		u32		reserved5[17];
+	};
 
 	/* USBMODE_EX: offset 0xc8 */
 	u32		usbmode_ex;	/* USB Device mode extension */
Signature · 1 line
2.26.0
#20 lore raw

Re: [PATCH 08/15] usb: ehci: avoid gcc-10 zero-length-bounds warning

Quoted 23 lines · Arnd Bergmann

Building ehci drivers with gcc-10 results in a number of warnings like when an zero-length array is accessed:

drivers/usb/host/ehci-hub.c: In function 'ehci_bus_suspend':
drivers/usb/host/ehci-hub.c:320:30: error: array subscript 14 is outside the bounds of an interior zero-length array 'u32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]

  320 |    u32 __iomem *hostpc_reg = &ehci->regs->hostpc[port];
      |                              ^~~~~~~~~~~~~~~~~~~~~~~~~
In file included from drivers/usb/host/ehci.h:273,
                 from drivers/usb/host/ehci-hcd.c:96:
include/linux/usb/ehci_def.h:186:7: note: while referencing 'hostpc'
  186 |  u32  hostpc[0]; /* HOSTPC extension */
      |       ^~~~~~
In file included from drivers/usb/host/ehci-hcd.c:305:
drivers/usb/host/ehci-hub.c: In function 'ehci_hub_control':
drivers/usb/host/ehci-hub.c:892:15: error: array subscript 256 is outside the bounds of an interior zero-length array 'u32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]
  892 |  hostpc_reg = &ehci->regs->hostpc[temp];
      |               ^~~~~~~~~~~~~~~~~~~~~~~~~
In file included from drivers/usb/host/ehci.h:273,
                 from drivers/usb/host/ehci-hcd.c:96:
include/linux/usb/ehci_def.h:186:7: note: while referencing 'hostpc'
  186 |  u32  hostpc[0]; /* HOSTPC extension */
      |       ^~~~~~
Quoted 2 lines

All these fields are colocated with reserved fields that I guess refer to the correct field length.

No, they don't.

Quoted 14 lines

Change the two struct definition to use an unnamed union to define both of these fields at the same location as the corresponding reserved fields.

---
 include/linux/usb/ehci_def.h | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
include/linux/usb/ehci_def.h +0 -0
index 78e006355557..8777d8e56ef2 100644

include/linux/usb/ehci_def.h struct ehci_regs { · line 131

Quoted 6 lines
include/linux/usb/ehci_def.h +2 -1
@@ -127,7 +127,8 @@ struct ehci_regs {
 #define FLAG_CF		(1<<0)		/* true: we'll support "high speed" */
 
 	/* PORTSC: offset 0x44 */
-	u32		port_status[0];	/* up to N_PORTS */
+	union {
+		u32		port_status[9];	/* up to N_PORTS */

This array can have up to 15 elements, meaning that it can extend out to offset 0x80.

include/linux/usb/ehci_def.h struct ehci_regs { · line 173

Quoted 13 lines
include/linux/usb/ehci_def.h +2 -1
 /* EHCI 1.1 addendum */
 #define PORTSC_SUSPEND_STS_ACK 0
 #define PORTSC_SUSPEND_STS_NYET 1
@@ -165,7 +166,8 @@ struct ehci_regs {
 #define PORT_CONNECT	(1<<0)		/* device connected */
 #define PORT_RWC_BITS   (PORT_CSC | PORT_PEC | PORT_OCC)
 
-	u32		reserved3[9];
+		u32		reserved3[9];
+	};
 
 	/* USBMODE: offset 0x68 */
 	u32		usbmode;	/* USB Device mode */

As you see, this next field actually lies inside the preceding array. It's not a real conflict; any hardware which supports the usbmode field uses only the first element of the port_status array.

I don't know how you want to handle this. Doing

#define usbmode port_status[9]

doesn't seem like a very good approach, but I can't think of anything better at the moment. Maybe just set the array size to 9, as you did, but with a comment explaining what's really going on.

include/linux/usb/ehci_def.h struct ehci_regs { · line 187

Quoted 7 lines
include/linux/usb/ehci_def.h +2 -1
@@ -181,11 +183,13 @@ struct ehci_regs {
  * PORTSCx
  */
 	/* HOSTPC: offset 0x84 */
-	u32		hostpc[0];	/* HOSTPC extension */
+	union {
+		u32		hostpc[17];	/* HOSTPC extension */

Likewise, this array can have up to 15 elements. In fact, it's the same size as the port_status array.

include/linux/usb/ehci_def.h struct ehci_regs { · line 195

Quoted 9 lines
include/linux/usb/ehci_def.h +2 -1
 #define HOSTPC_PHCD	(1<<22)		/* Phy clock disable */
 #define HOSTPC_PSPD	(3<<25)		/* Port speed detection */
 
-	u32		reserved5[17];
+		u32		reserved5[17];
+	};
 
 	/* USBMODE_EX: offset 0xc8 */
 	u32		usbmode_ex;	/* USB Device mode extension */

Alan Stern

#21 lore raw

Re: [PATCH 08/15] usb: ehci: avoid gcc-10 zero-length-bounds warning

Quoted 5 lines · Alan Stern

On Thu, 30 Apr 2020, Arnd Bergmann wrote:

No, they don't.

Quoted 7 lines

include/linux/usb/ehci_def.h struct ehci_regs { · line 131

Quoted 4 lines
      /* PORTSC: offset 0x44 */
include/linux/usb/ehci_def.h +2 -1
-     u32             port_status[0]; /* up to N_PORTS */
+     union {
+             u32             port_status[9]; /* up to N_PORTS */

This array can have up to 15 elements, meaning that it can extend out to offset 0x80.

Ok, thanks for the clarification!

Quoted 18 lines

include/linux/usb/ehci_def.h struct ehci_regs { · line 173

Quoted 13 lines

/* EHCI 1.1 addendum */
#define PORTSC_SUSPEND_STS_ACK 0
#define PORTSC_SUSPEND_STS_NYET 1

include/linux/usb/ehci_def.h +2 -1
@@ -165,7 +166,8 @@ struct ehci_regs {
 #define PORT_CONNECT (1<<0)          /* device connected */
 #define PORT_RWC_BITS   (PORT_CSC | PORT_PEC | PORT_OCC)
 
-     u32             reserved3[9];
+             u32             reserved3[9];
+     };
 
      /* USBMODE: offset 0x68 */
      u32             usbmode;        /* USB Device mode */

As you see, this next field actually lies inside the preceding array. It's not a real conflict; any hardware which supports the usbmode field uses only the first element of the port_status array.

Quoted 7 lines

I don't know how you want to handle this. Doing

#define usbmode port_status[9]

doesn't seem like a very good approach, but I can't think of anything better at the moment. Maybe just set the array size to 9, as you did, but with a comment explaining what's really going on.

The easiest change would be to use an anonymous struct like this

        /* PORTSC: offset 0x44 */
        union {
                u32             port_status[15]; /* up to N_PORTS */
/* EHCI 1.1 addendum */
#define PORTSC_SUSPEND_STS_ACK 0
...
#define PORT_RWC_BITS   (PORT_CSC | PORT_PEC | PORT_OCC)
                struct {
                        u32             reserved3[9];

        /* USBMODE: offset 0x68 */
                        u32             usbmode;        /* USB Device mode */
#define USBMODE_SDIS    (1<<3)          /* Stream disable */
#define USBMODE_BE      (1<<2)          /* BE/LE endianness select */
#define USBMODE_CM_HC   (3<<0)          /* host controller mode */
#define USBMODE_CM_IDLE (0<<0)          /* idle state */
                        u32             reserved4[5];
                };
        };
        u32             reserved5;

It doesn't really improve readability, but it should correctly reflect the layout as you described it.

Arnd

#22 lore raw

Re: [PATCH 08/15] usb: ehci: avoid gcc-10 zero-length-bounds warning

Quoted 61 lines · Arnd Bergmann
Quoted 5 lines · Alan Stern

On Thu, 30 Apr 2020, Arnd Bergmann wrote:

No, they don't.

Quoted 7 lines

include/linux/usb/ehci_def.h struct ehci_regs { · line 131

Quoted 4 lines
      /* PORTSC: offset 0x44 */
include/linux/usb/ehci_def.h +2 -1
-     u32             port_status[0]; /* up to N_PORTS */
+     union {
+             u32             port_status[9]; /* up to N_PORTS */

This array can have up to 15 elements, meaning that it can extend out to offset 0x80.

Ok, thanks for the clarification!

Quoted 18 lines

include/linux/usb/ehci_def.h struct ehci_regs { · line 173

Quoted 13 lines

/* EHCI 1.1 addendum */
#define PORTSC_SUSPEND_STS_ACK 0
#define PORTSC_SUSPEND_STS_NYET 1

include/linux/usb/ehci_def.h +2 -1
@@ -165,7 +166,8 @@ struct ehci_regs {
 #define PORT_CONNECT (1<<0)          /* device connected */
 #define PORT_RWC_BITS   (PORT_CSC | PORT_PEC | PORT_OCC)
 
-     u32             reserved3[9];
+             u32             reserved3[9];
+     };
 
      /* USBMODE: offset 0x68 */
      u32             usbmode;        /* USB Device mode */

As you see, this next field actually lies inside the preceding array. It's not a real conflict; any hardware which supports the usbmode field uses only the first element of the port_status array.

Quoted 7 lines

I don't know how you want to handle this. Doing

#define usbmode port_status[9]

doesn't seem like a very good approach, but I can't think of anything better at the moment. Maybe just set the array size to 9, as you did, but with a comment explaining what's really going on.

The easiest change would be to use an anonymous struct like this

        /* PORTSC: offset 0x44 */
        union {
                u32             port_status[15]; /* up to N_PORTS */
/* EHCI 1.1 addendum */
#define PORTSC_SUSPEND_STS_ACK 0
...
#define PORT_RWC_BITS   (PORT_CSC | PORT_PEC | PORT_OCC)
                struct {
                        u32             reserved3[9];

        /* USBMODE: offset 0x68 */
                        u32             usbmode;        /* USB Device mode */
#define USBMODE_SDIS    (1<<3)          /* Stream disable */
#define USBMODE_BE      (1<<2)          /* BE/LE endianness select */
#define USBMODE_CM_HC   (3<<0)          /* host controller mode */
#define USBMODE_CM_IDLE (0<<0)          /* idle state */
Quoted 7 lines
                        u32             reserved4[5];
                };
        };
        u32             reserved5;

It doesn't really improve readability, but it should correctly reflect the layout as you described it.

Sounds good. Please go ahead and update the patch.

Alan Stern

#23 lore raw

PATCH 9/15 udf: avoid gcc-10 zero-length-bounds warnings

In reply to Arnd Bergmann

gcc-10 warns about writes to the empty freeSpaceTable[] array, with many instances like:

fs/udf/balloc.c: In function 'udf_bitmap_new_block':
fs/udf/balloc.c:101:36: error: array subscript 65535 is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]

  101 |  le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
      |                ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~
In file included from fs/udf/udfdecl.h:7,
                 from fs/udf/balloc.c:22:
fs/udf/ecma_167.h:363:11: note: while referencing 'freeSpaceTable'
  363 |  __le32   freeSpaceTable[0];
      |           ^~~~~~~~~~~~~~

These can all be avoided by using a flexible array member instead.

Another warning is a bit more obscure:

fs/udf/super.c: In function 'udf_count_free':
fs/udf/super.c:2521:26: warning: array subscript '(<unknown>) + 4294967295' is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Wzero-length-bounds]
2521 | lvid->freeSpaceTable[part]);

Work around this one by changing the array access to equivalent pointer arithmetic, as there cannot be multiple flexible-array members in a single struct.

---
 fs/udf/ecma_167.h | 2 +-
 fs/udf/super.c    | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/udf/ecma_167.h +1 -1
index 736ebc5dc441..14ffe27342bc 100644
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
 	uint8_t			logicalVolContentsUse[32];
 	__le32			numOfPartitions;
 	__le32			lengthOfImpUse;
-	__le32			freeSpaceTable[0];
 	__le32			sizeTable[0];
 	uint8_t			impUse[0];
+	__le32			freeSpaceTable[];
 } __packed;
 
 /* Integrity Type (ECMA 167r3 3/10.10.3) */
fs/udf/super.c +1 -1
index f747bf72edbe..379867888c36 100644
@@ -117,7 +117,7 @@ struct logicalVolIntegrityDescImpUse *udf_sb_lvidiu(struct super_block *sb)
 	}
 	/* The offset is to skip freeSpaceTable and sizeTable arrays */
 	offset = partnum * 2 * sizeof(uint32_t);
-	return (struct logicalVolIntegrityDescImpUse *)&(lvid->impUse[offset]);
+	return (struct logicalVolIntegrityDescImpUse *)(lvid->impUse + offset);
 }
 
 /* UDF filesystem type */
Signature · 1 line
2.26.0
#24 lore raw

Re: [PATCH 09/15] udf: avoid gcc-10 zero-length-bounds warnings

Quoted 5 lines · On Thursday 30 April 2020 23:30:51 Arnd Bergmann wrote:

gcc-10 warns about writes to the empty freeSpaceTable[] array, with many instances like:

fs/udf/balloc.c: In function 'udf_bitmap_new_block':
fs/udf/balloc.c:101:36: error: array subscript 65535 is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]

Quoted 7 lines
  101 |  le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
      |                ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~
In file included from fs/udf/udfdecl.h:7,
                 from fs/udf/balloc.c:22:
fs/udf/ecma_167.h:363:11: note: while referencing 'freeSpaceTable'
  363 |  __le32   freeSpaceTable[0];
      |           ^~~~~~~~~~~~~~

Hi Arnd! This looks like a false-positive warning.

Quoted 11 lines

These can all be avoided by using a flexible array member instead.

Another warning is a bit more obscure:

fs/udf/super.c: In function 'udf_count_free':
fs/udf/super.c:2521:26: warning: array subscript '(<unknown>) + 4294967295' is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Wzero-length-bounds]
2521 | lvid->freeSpaceTable[part]);

Work around this one by changing the array access to equivalent pointer arithmetic, as there cannot be multiple flexible-array members in a single struct.

Well, this code uses GNU extension for zero-length arrays because it was written in pre-C99 era when C99 flexible arrays did not exist yet.

Quoted 11 lines
---
 fs/udf/ecma_167.h | 2 +-
 fs/udf/super.c    | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/udf/ecma_167.h +0 -0
index 736ebc5dc441..14ffe27342bc 100644

fs/udf/ecma_167.h struct logicalVolIntegrityDesc { · line 365

Quoted 7 lines
fs/udf/ecma_167.h +1 -1
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
 	uint8_t			logicalVolContentsUse[32];
 	__le32			numOfPartitions;
 	__le32			lengthOfImpUse;
-	__le32			freeSpaceTable[0];
 	__le32			sizeTable[0];
 	uint8_t			impUse[0];
+	__le32			freeSpaceTable[];

Please do not change order of members in these structures. Order is strictly defined by ECMA 167 standard and changing them you would just confuse reader. In LVID is free space table before size table.

If you do not like GNU C extension for zero-length arrays then just replace it by standard C99 flexible arrays. I think that there is no reason to not use standard C99 language constructions, just nobody had motivation or time to change (working) code.

Also this file is semi-synchronized with udftools project in which I already replaced all GNU C zero-length arrays by C99 flexible arrays.

You can take inspiration what I did with logicalVolIntegrityDesc:
https://github.com/pali/udftools/commit/f851d84478ce881d516a76018745fa163f803880#diff-1e1a5b89f620d380f22b973f9449aeaeL381-R384

Anyway, if you have a better idea what to do with such on-disk structure and how to represent it in C struct syntax, let me know as it could be updated also in udftools project.

Quoted 19 lines
fs/udf/ecma_167.h +0 -0
 } __packed;
 
 /* Integrity Type (ECMA 167r3 3/10.10.3) */
fs/udf/super.c +1 -1
index f747bf72edbe..379867888c36 100644
@@ -117,7 +117,7 @@ struct logicalVolIntegrityDescImpUse *udf_sb_lvidiu(struct super_block *sb)
 	}
 	/* The offset is to skip freeSpaceTable and sizeTable arrays */
 	offset = partnum * 2 * sizeof(uint32_t);
-	return (struct logicalVolIntegrityDescImpUse *)&(lvid->impUse[offset]);
+	return (struct logicalVolIntegrityDescImpUse *)(lvid->impUse + offset);
 }
 
 /* UDF filesystem type */
Signature · 1 line
2.26.0
#25 lore raw

Re: [PATCH 09/15] udf: avoid gcc-10 zero-length-bounds warnings

Quoted 15 lines · Pali Rohár
Quoted 5 lines · On Thursday 30 April 2020 23:30:51 Arnd Bergmann wrote:

gcc-10 warns about writes to the empty freeSpaceTable[] array, with many instances like:

fs/udf/balloc.c: In function 'udf_bitmap_new_block':
fs/udf/balloc.c:101:36: error: array subscript 65535 is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]

Quoted 7 lines
  101 |  le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
      |                ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~
In file included from fs/udf/udfdecl.h:7,
                 from fs/udf/balloc.c:22:
fs/udf/ecma_167.h:363:11: note: while referencing 'freeSpaceTable'
  363 |  __le32   freeSpaceTable[0];
      |           ^~~~~~~~~~~~~~
Quoted 1 line

Hi Arnd! This looks like a false-positive warning.

Right, sorry for not making that clearer in the changelog.

Quoted 13 lines
Quoted 11 lines

These can all be avoided by using a flexible array member instead.

Another warning is a bit more obscure:

fs/udf/super.c: In function 'udf_count_free':
fs/udf/super.c:2521:26: warning: array subscript '(<unknown>) + 4294967295' is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Wzero-length-bounds]
2521 | lvid->freeSpaceTable[part]);

Work around this one by changing the array access to equivalent pointer arithmetic, as there cannot be multiple flexible-array members in a single struct.

Quoted 12 lines

Review struct logicalVolIntegrityDesc { · line 365

Quoted 8 lines
(quoted diff) +1 -1
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
      uint8_t                 logicalVolContentsUse[32];
      __le32                  numOfPartitions;
      __le32                  lengthOfImpUse;
-     __le32                  freeSpaceTable[0];
      __le32                  sizeTable[0];
      uint8_t                 impUse[0];
+     __le32                  freeSpaceTable[];

Please do not change order of members in these structures. Order is strictly defined by ECMA 167 standard and changing them you would just confuse reader. In LVID is free space table before size table.

Ok

Quoted 4 lines

If you do not like GNU C extension for zero-length arrays then just replace it by standard C99 flexible arrays. I think that there is no reason to not use standard C99 language constructions, just nobody had motivation or time to change (working) code.

No, the problem is that only the last member can be a flexible array, so when impUse[] is the last member, freeSpaceTable has to be a zero length array.

[]> Also this file is semi-synchronized with udftools project in which I

Quoted 4 lines

already replaced all GNU C zero-length arrays by C99 flexible arrays.

You can take inspiration what I did with logicalVolIntegrityDesc:
https://github.com/pali/udftools/commit/f851d84478ce881d516a76018745fa163f803880#diff-1e1a5b89f620d380f22b973f9449aeaeL381-R384

Right, this is likely the best workaround.

Quoted 3 lines

Anyway, if you have a better idea what to do with such on-disk structure and how to represent it in C struct syntax, let me know as it could be updated also in udftools project.

The trick I used for impUse[] would also work for freeSpaceTable[] to avoid the gcc warning, it's still not great, but maybe you like this better:

arnd@threadripper:~/arm-soc$ git diff
fs/udf/balloc.c +0 -0
index 02f03fadb75b..666d022eb00b 100644
@@ -98,7 +98,7 @@ static void udf_add_free_space(struct super_block

*sb, u16 partition, u32 cnt)

                return;

        lvid = (struct logicalVolIntegrityDesc *)sbi->s_lvid_bh->b_data;
-       le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
+       le32_add_cpu(lvid->freeSpaceTable + partition, cnt);
        udf_updated_lvid(sb);

}

fs/udf/ecma_167.h +1 -1
index 14ffe27342bc..215d97d7edc4 100644
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
        uint8_t                 logicalVolContentsUse[32];
        __le32                  numOfPartitions;
        __le32                  lengthOfImpUse;
        __le32                  freeSpaceTable[0];
        __le32                  sizeTable[0];
-       uint8_t                 impUse[0];
+       uint8_t                 impUse[];
 } __packed;
 
 /* Integrity Type (ECMA 167r3 3/10.10.3) */
fs/udf/super.c +2 -2
index 379867888c36..a1fc51c2261e 100644
@@ -2517,8 +2517,8 @@ static unsigned int udf_count_free(struct super_block *sb)
                        (struct logicalVolIntegrityDesc *)
                        sbi->s_lvid_bh->b_data;
                if (le32_to_cpu(lvid->numOfPartitions) > part) {
-                       accum = le32_to_cpu(
-                                       lvid->freeSpaceTable[part]);
+                       accum = le32_to_cpup(
+                                       (lvid->freeSpaceTable + part));
                        if (accum == 0xFFFFFFFF)
                                accum = 0;
                }

This version could easily be backported to stable kernels to let them be compiled with gcc-10, and then synchronizing with the udftools version of the header needs additional changes on top, which do not need to be backported.

Arnd

#26 lore raw

Re: [PATCH 09/15] udf: avoid gcc-10 zero-length-bounds warnings

Quoted 31 lines · Arnd Bergmann
Quoted 4 lines · Pali Rohár
Quoted 4 lines
fs/udf/ecma_167.h +0 -0
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
      uint8_t                 logicalVolContentsUse[32];
      __le32                  numOfPartitions;
      __le32                  lengthOfImpUse;
Quoted 8 lines

Review struct logicalVolIntegrityDesc { · line 365

Quoted 4 lines
(quoted diff) +1 -1
-     __le32                  freeSpaceTable[0];
      __le32                  sizeTable[0];
      uint8_t                 impUse[0];
+     __le32                  freeSpaceTable[];

Please do not change order of members in these structures. Order is strictly defined by ECMA 167 standard and changing them you would just confuse reader. In LVID is free space table before size table.

Ok

Quoted 4 lines

If you do not like GNU C extension for zero-length arrays then just replace it by standard C99 flexible arrays. I think that there is no reason to not use standard C99 language constructions, just nobody had motivation or time to change (working) code.

No, the problem is that only the last member can be a flexible array, so when impUse[] is the last member, freeSpaceTable has to be a zero length array.

[]> Also this file is semi-synchronized with udftools project in which I

Quoted 4 lines

already replaced all GNU C zero-length arrays by C99 flexible arrays.

You can take inspiration what I did with logicalVolIntegrityDesc:
https://github.com/pali/udftools/commit/f851d84478ce881d516a76018745fa163f803880#diff-1e1a5b89f620d380f22b973f9449aeaeL381-R384

Quoted 8 lines

Right, this is likely the best workaround.

Quoted 3 lines

Anyway, if you have a better idea what to do with such on-disk structure and how to represent it in C struct syntax, let me know as it could be updated also in udftools project.

The trick I used for impUse[] would also work for freeSpaceTable[] to avoid the gcc warning, it's still not great, but maybe you like this better:

I like Pali's version somewhat better because whenever I look at several (obviously flexible) arrays in one struct, I start wondering what's going on. So let's not define members of struct whose offset we actually don't know (and thus we cannot sanely use them anyway).

Honza

Quoted 54 lines
arnd@threadripper:~/arm-soc$ git diff
fs/udf/balloc.c +1 -1
index 02f03fadb75b..666d022eb00b 100644
@@ -98,7 +98,7 @@ static void udf_add_free_space(struct super_block
 *sb, u16 partition, u32 cnt)
                return;
 
        lvid = (struct logicalVolIntegrityDesc *)sbi->s_lvid_bh->b_data;
-       le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
+       le32_add_cpu(lvid->freeSpaceTable + partition, cnt);
        udf_updated_lvid(sb);
 }
fs/udf/ecma_167.h +1 -1
index 14ffe27342bc..215d97d7edc4 100644
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
        uint8_t                 logicalVolContentsUse[32];
        __le32                  numOfPartitions;
        __le32                  lengthOfImpUse;
        __le32                  freeSpaceTable[0];
        __le32                  sizeTable[0];
-       uint8_t                 impUse[0];
+       uint8_t                 impUse[];
 } __packed;
 
 /* Integrity Type (ECMA 167r3 3/10.10.3) */
fs/udf/super.c +2 -2
index 379867888c36..a1fc51c2261e 100644
@@ -2517,8 +2517,8 @@ static unsigned int udf_count_free(struct super_block *sb)
                        (struct logicalVolIntegrityDesc *)
                        sbi->s_lvid_bh->b_data;
                if (le32_to_cpu(lvid->numOfPartitions) > part) {
-                       accum = le32_to_cpu(
-                                       lvid->freeSpaceTable[part]);
+                       accum = le32_to_cpup(
+                                       (lvid->freeSpaceTable + part));
                        if (accum == 0xFFFFFFFF)
                                accum = 0;
                }

This version could easily be backported to stable kernels to let them be compiled with gcc-10, and then synchronizing with the udftools version of the header needs additional changes on top, which do not need to be backported.

Arnd

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
Acked-by
#27 lore raw

Re: [PATCH 09/15] udf: avoid gcc-10 zero-length-bounds warnings

In reply to Arnd Bergmann
Quoted 46 lines · On Friday 01 May 2020 22:30:27 Arnd Bergmann wrote:
Quoted 15 lines · Pali Rohár
Quoted 5 lines · On Thursday 30 April 2020 23:30:51 Arnd Bergmann wrote:

gcc-10 warns about writes to the empty freeSpaceTable[] array, with many instances like:

fs/udf/balloc.c: In function 'udf_bitmap_new_block':
fs/udf/balloc.c:101:36: error: array subscript 65535 is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Werror=zero-length-bounds]

Quoted 7 lines
  101 |  le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
      |                ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~
In file included from fs/udf/udfdecl.h:7,
                 from fs/udf/balloc.c:22:
fs/udf/ecma_167.h:363:11: note: while referencing 'freeSpaceTable'
  363 |  __le32   freeSpaceTable[0];
      |           ^~~~~~~~~~~~~~
Quoted 1 line

Hi Arnd! This looks like a false-positive warning.

Right, sorry for not making that clearer in the changelog.

Quoted 13 lines
Quoted 11 lines

These can all be avoided by using a flexible array member instead.

Another warning is a bit more obscure:

fs/udf/super.c: In function 'udf_count_free':
fs/udf/super.c:2521:26: warning: array subscript '(<unknown>) + 4294967295' is outside the bounds of an interior zero-length array '__le32[0]' {aka 'unsigned int[0]'} [-Wzero-length-bounds]
2521 | lvid->freeSpaceTable[part]);

Work around this one by changing the array access to equivalent pointer arithmetic, as there cannot be multiple flexible-array members in a single struct.

Quoted 12 lines

fs/udf/ecma_167.h struct logicalVolIntegrityDesc { · line 365

Quoted 8 lines
fs/udf/ecma_167.h +1 -1
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
      uint8_t                 logicalVolContentsUse[32];
      __le32                  numOfPartitions;
      __le32                  lengthOfImpUse;
-     __le32                  freeSpaceTable[0];
      __le32                  sizeTable[0];
      uint8_t                 impUse[0];
+     __le32                  freeSpaceTable[];

Please do not change order of members in these structures. Order is strictly defined by ECMA 167 standard and changing them you would just confuse reader. In LVID is free space table before size table.

Quoted 8 lines

Ok

Quoted 4 lines

If you do not like GNU C extension for zero-length arrays then just replace it by standard C99 flexible arrays. I think that there is no reason to not use standard C99 language constructions, just nobody had motivation or time to change (working) code.

No, the problem is that only the last member can be a flexible array,

I know, that is why I replaced those 3 zero-length arrays by just one flexible array in udftools project.

Quoted 54 lines

so when impUse[] is the last member, freeSpaceTable has to be a zero length array.

Quoted 5 lines

Also this file is semi-synchronized with udftools project in which I already replaced all GNU C zero-length arrays by C99 flexible arrays.

You can take inspiration what I did with logicalVolIntegrityDesc:
https://github.com/pali/udftools/commit/f851d84478ce881d516a76018745fa163f803880#diff-1e1a5b89f620d380f22b973f9449aeaeL381-R384

Right, this is likely the best workaround.

Quoted 3 lines

Anyway, if you have a better idea what to do with such on-disk structure and how to represent it in C struct syntax, let me know as it could be updated also in udftools project.

The trick I used for impUse[] would also work for freeSpaceTable[] to avoid the gcc warning, it's still not great, but maybe you like this better:

arnd@threadripper:~/arm-soc$ git diff
fs/udf/balloc.c +1 -1
index 02f03fadb75b..666d022eb00b 100644
@@ -98,7 +98,7 @@ static void udf_add_free_space(struct super_block
 *sb, u16 partition, u32 cnt)
                return;
 
        lvid = (struct logicalVolIntegrityDesc *)sbi->s_lvid_bh->b_data;
-       le32_add_cpu(&lvid->freeSpaceTable[partition], cnt);
+       le32_add_cpu(lvid->freeSpaceTable + partition, cnt);
        udf_updated_lvid(sb);
 }
fs/udf/ecma_167.h +1 -1
index 14ffe27342bc..215d97d7edc4 100644
@@ -360,9 +360,9 @@ struct logicalVolIntegrityDesc {
        uint8_t                 logicalVolContentsUse[32];
        __le32                  numOfPartitions;
        __le32                  lengthOfImpUse;
        __le32                  freeSpaceTable[0];
        __le32                  sizeTable[0];
-       uint8_t                 impUse[0];
+       uint8_t                 impUse[];
 } __packed;
 
 /* Integrity Type (ECMA 167r3 3/10.10.3) */
fs/udf/super.c +0 -0
index 379867888c36..a1fc51c2261e 100644

fs/udf/super.c static unsigned int udf_count_free(struct super_block *sb) · line 2524

Quoted 10 lines
fs/udf/super.c +2 -2
@@ -2517,8 +2517,8 @@ static unsigned int udf_count_free(struct super_block *sb)
                        (struct logicalVolIntegrityDesc *)
                        sbi->s_lvid_bh->b_data;
                if (le32_to_cpu(lvid->numOfPartitions) > part) {
-                       accum = le32_to_cpu(
-                                       lvid->freeSpaceTable[part]);
+                       accum = le32_to_cpup(
+                                       (lvid->freeSpaceTable + part));
                        if (accum == 0xFFFFFFFF)
                                accum = 0;
                }

This is much better as it does not change order of members in LVID structure. I'm fine with it.

Quoted 2 lines

This version could easily be backported to stable kernels to let them be compiled with gcc-10

I do not know what triggers that false-positive warning. But if you think that this change is enough to "hide" that warning, you can add my

For sure it is better to have just small changes needed for backporting.

Quoted 3 lines

and then synchronizing with the udftools version of
the header needs additional changes on top, which do not need to be backported.

Both header files (ECMA and OSTA) should be in-sync with udftools since
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=614644676394951e73194ea96b3f026c1adf5443
Differences in kernel code are: usage of zero-length array members, usage of integer types and usage of structure attributes.

If you are planning in future to do some changes in those ECMA or OSTA header files, please send updates also for udftools. So we will have header files synchronized as much as possible.

#28 lore raw

PATCH 10/15 hpfs: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

gcc-10 now warns about accessing members of a zero-length array field:

fs/hpfs/anode.c: In function 'hpfs_add_sector_to_btree':
fs/hpfs/anode.c:209:27: error: array subscript 0 is outside the bounds of an interior zero-length array 'struct bplus_internal_node[0]' [-Werror=zero-length-bounds]

  209 |    anode->btree.u.internal[0].down = cpu_to_le32(a);
      |    ~~~~~~~~~~~~~~~~~~~~~~~^~~
In file included from fs/hpfs/hpfs_fn.h:26,
                 from fs/hpfs/anode.c:10:

fs/hpfs/hpfs.h:411:32: note: while referencing 'internal'
411 | struct bplus_internal_node internal[0]; /* (internal) 2-word entries giving
| ^~~~~~~~

It would be nice to make it a flexible-array member, but that is not possible inside of a union, so work around the warnings by adding a temporary pointer that hides the problem.

---
 fs/hpfs/anode.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
fs/hpfs/anode.c +5 -2
index c14c9a035ee0..55788ca7969f 100644
@@ -200,14 +200,17 @@ secno hpfs_add_sector_to_btree(struct super_block *s, secno node, int fnod, unsi
 		brelse(bh);
 		a = na;
 		if ((new_anode = hpfs_alloc_anode(s, a, &na, &bh))) {
+			struct bplus_internal_node *internal;
+
 			anode = new_anode;
 			/*anode->up = cpu_to_le32(up != -1 ? up : ra);*/
 			anode->btree.flags |= BP_internal;
 			anode->btree.n_used_nodes = 1;
 			anode->btree.n_free_nodes = 59;
 			anode->btree.first_free = cpu_to_le16(16);
-			anode->btree.u.internal[0].down = cpu_to_le32(a);
-			anode->btree.u.internal[0].file_secno = cpu_to_le32(-1);
+			internal = anode->btree.u.internal;
+			internal->down = cpu_to_le32(a);
+			internal->file_secno = cpu_to_le32(-1);
 			mark_buffer_dirty(bh);
 			brelse(bh);
 			if ((anode = hpfs_map_anode(s, a, &bh))) {
Signature · 1 line
2.26.0
#29 lore raw

PATCH 11/15 omfs: avoid gcc-10 stringop-overflow warning

In reply to Arnd Bergmann

gcc-10 correctly points out a memcpy beyond the end of the structure it gets copied into:

fs/omfs/file.c: In function 'omfs_grow_extent':
cc1: warning: writing 16 bytes into a region of size 0 [-Wstringop-overflow=] In file included from fs/omfs/omfs.h:8,

                 from fs/omfs/file.c:11:

fs/omfs/omfs_fs.h:80:27: note: at offset 0 to object 'e_entry' with size 16 declared here
80 | struct omfs_extent_entry e_entry; /* start of extent entries */
| ^~~~~~~

This is not a bug however, as the file system layout contains an array of extent entries instead of just a single one that was part of the structure definition.

Changing the definition to use a c99 flexible array member makes makes the code match the actual layout, which avoids the warning and an odd '+1' in the calculation of the number of extent entries.

Fixes 8f09e98768c1 ("omfs: add file routines")
---
 fs/omfs/file.c    | 12 ++++++------
 fs/omfs/omfs_fs.h |  2 +-
 2 files changed, 7 insertions(+), 7 deletions(-)
fs/omfs/file.c +6 -6
index d7b5f09d298c..94344386c6ff 100644
@@ -14,7 +14,7 @@ static u32 omfs_max_extents(struct omfs_sb_info *sbi, int offset)
 {
 	return (sbi->s_sys_blocksize - offset -
 		sizeof(struct omfs_extent)) /
-		sizeof(struct omfs_extent_entry) + 1;
+		sizeof(struct omfs_extent_entry);
 }
 
 void omfs_make_empty_table(struct buffer_head *bh, int offset)
@@ -24,8 +24,8 @@ void omfs_make_empty_table(struct buffer_head *bh, int offset)
 	oe->e_next = ~cpu_to_be64(0ULL);
 	oe->e_extent_count = cpu_to_be32(1),
 	oe->e_fill = cpu_to_be32(0x22),
-	oe->e_entry.e_cluster = ~cpu_to_be64(0ULL);
-	oe->e_entry.e_blocks = ~cpu_to_be64(0ULL);
+	oe->e_entry[0].e_cluster = ~cpu_to_be64(0ULL);
+	oe->e_entry[0].e_blocks = ~cpu_to_be64(0ULL);
 }
 
 int omfs_shrink_inode(struct inode *inode)
@@ -68,7 +68,7 @@ int omfs_shrink_inode(struct inode *inode)
 
 		last = next;
 		next = be64_to_cpu(oe->e_next);
-		entry = &oe->e_entry;
+		entry = oe->e_entry;
 
 		/* ignore last entry as it is the terminator */
 		for (; extent_count > 1; extent_count--) {
@@ -117,7 +117,7 @@ static int omfs_grow_extent(struct inode *inode, struct omfs_extent *oe,
 			u64 *ret_block)
 {
 	struct omfs_extent_entry *terminator;
-	struct omfs_extent_entry *entry = &oe->e_entry;
+	struct omfs_extent_entry *entry = oe->e_entry;
 	struct omfs_sb_info *sbi = OMFS_SB(inode->i_sb);
 	u32 extent_count = be32_to_cpu(oe->e_extent_count);
 	u64 new_block = 0;
@@ -245,7 +245,7 @@ static int omfs_get_block(struct inode *inode, sector_t block,
 
 		extent_count = be32_to_cpu(oe->e_extent_count);
 		next = be64_to_cpu(oe->e_next);
-		entry = &oe->e_entry;
+		entry = oe->e_entry;
 
 		if (extent_count > max_extents)
 			goto out_brelse;
fs/omfs/omfs_fs.h +1 -1
index caecb3d5a344..1ff6b9e41297 100644
@@ -77,7 +77,7 @@ struct omfs_extent {
 	__be64 e_next;			/* next extent table location */
 	__be32 e_extent_count;		/* total # extents in this table */
 	__be32 e_fill;
-	struct omfs_extent_entry e_entry;	/* start of extent entries */
+	struct omfs_extent_entry e_entry[];	/* start of extent entries */
 };
 
 #endif
Signature · 1 line
2.26.0
#30 lore raw

PATCH 12/15 media: s5k5baf: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

gcc-10 warns about accessing a zero-length struct member:

drivers/media/i2c/s5k5baf.c: In function 's5k5baf_load_setfile':
drivers/media/i2c/s5k5baf.c:390:13: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct <anonymous>[0]' [-Wzero-length-bounds]
390 | if (f->seq[i].offset + d <= end)

      |       ~~~~~~^~~

This should really be a flexible-array member, but the structure already has one. I experimentally confirmed that swapping the two avoids the warning, as the 'data[]' array is not accessed like this.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 7d459937dc09 ("[media] Add driver for Samsung S5K5BAF camera sensor")
---
 drivers/media/i2c/s5k5baf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/media/i2c/s5k5baf.c +2 -2
index 42584a088273..0b1ddedcf7dc 100644
@@ -277,11 +277,11 @@ enum {
 
 struct s5k5baf_fw {
 	u16 count;
+	u16 data[0];
 	struct {
 		u16 id;
 		u16 offset;
-	} seq[0];
-	u16 data[];
+	} seq[];
 };
 
 struct s5k5baf {
Signature · 1 line
2.26.0
#31 lore raw

Re: [PATCH 12/15] media: s5k5baf: avoid gcc-10 zero-length-bounds warning

Hi Arnd,

Quoted 27 lines · Arnd Bergmann

gcc-10 warns about accessing a zero-length struct member:

drivers/media/i2c/s5k5baf.c: In function 's5k5baf_load_setfile':
drivers/media/i2c/s5k5baf.c:390:13: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'struct <anonymous>[0]' [-Wzero-length-bounds]
390 | if (f->seq[i].offset + d <= end)

      |       ~~~~~~^~~

This should really be a flexible-array member, but the structure already has one. I experimentally confirmed that swapping the two avoids the warning, as the 'data[]' array is not accessed like this.

Fixes 3ba225b506a2 ("treewide: Replace zero-length array with flexible-array member")
Fixes 7d459937dc09 ("[media] Add driver for Samsung S5K5BAF camera sensor")
---
 drivers/media/i2c/s5k5baf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/media/i2c/s5k5baf.c +0 -0
index 42584a088273..0b1ddedcf7dc 100644
@@ -277,11 +277,11 @@ enum {
 
 struct s5k5baf_fw {
 	u16 count;

drivers/media/i2c/s5k5baf.c enum { · line 287

Quoted 10 lines
drivers/media/i2c/s5k5baf.c +2 -2
+	u16 data[0];
 	struct {
 		u16 id;
 		u16 offset;
-	} seq[0];
-	u16 data[];
+	} seq[];
 };
 
 struct s5k5baf {

The treewide patch is an experimental change and, as this change only applies to my -next tree, I will carry this patch in it, so other people don't have to worry about this at all.

Thank you
--
Gustavo

#32 lore raw

PATCH 13/15 scsi: sas: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

Two files access the zero-length resp_data[] array, which now causes a compiler warning:

drivers/scsi/aic94xx/aic94xx_tmf.c: In function 'asd_get_tmf_resp_tasklet':
drivers/scsi/aic94xx/aic94xx_tmf.c:291:22: warning: array subscript 3 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]

  291 |   res = ru->resp_data[3];
      |         ~~~~~~~~~~~~~^~~
In file included from include/scsi/libsas.h:15,
                 from drivers/scsi/aic94xx/aic94xx.h:16,
                 from drivers/scsi/aic94xx/aic94xx_tmf.c:11:
include/scsi/sas.h:557:9: note: while referencing 'resp_data'
  557 |  u8     resp_data[0];
      |         ^~~~~~~~~
drivers/scsi/libsas/sas_task.c: In function 'sas_ssp_task_response':
drivers/scsi/libsas/sas_task.c:21:30: warning: array subscript 3 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]
   21 |   tstat->stat = iu->resp_data[3];
      |                 ~~~~~~~~~~~~~^~~

In file included from include/scsi/scsi_transport_sas.h:8,
from drivers/scsi/libsas/sas_internal.h:14, from drivers/scsi/libsas/sas_task.c:3:
include/scsi/sas.h:557:9: note: while referencing 'resp_data'

  557 |  u8     resp_data[0];
      |         ^~~~~~~~~

This should really be a flexible-array member, but the structure already has such a member, swapping it out with sense_data[] would cause many more warnings elsewhere.

As a workaround, add a temporary pointer that can be accessed without a warning.

Fixes 2908d778ab3e ("[SCSI] aic94xx: new driver")
Fixes 366ca51f30de ("[SCSI] libsas: abstract STP task status into a function")
---
 drivers/scsi/aic94xx/aic94xx_tmf.c | 4 +++-
 drivers/scsi/libsas/sas_task.c     | 3 ++-
 2 files changed, 5 insertions(+), 2 deletions(-)
drivers/scsi/aic94xx/aic94xx_tmf.c +3 -1
index f814026f26fa..a3139f9766c8 100644
@@ -269,6 +269,7 @@ static int asd_get_tmf_resp_tasklet(struct asd_ascb *ascb,
 	struct ssp_frame_hdr *fh;
 	struct ssp_response_iu   *ru;
 	int res = TMF_RESP_FUNC_FAILED;
+	u8 *resp;
 
 	ASD_DPRINTK("tmf resp tasklet\n");
 
@@ -287,8 +288,9 @@ static int asd_get_tmf_resp_tasklet(struct asd_ascb *ascb,
 	fh = edb->vaddr + 16;
 	ru = edb->vaddr + 16 + sizeof(*fh);
 	res = ru->status;
+	resp = ru->resp_data;
 	if (ru->datapres == 1)	  /* Response data present */
-		res = ru->resp_data[3];
+		res = resp[3];
 #if 0
 	ascb->tag = fh->tag;
 #endif
drivers/scsi/libsas/sas_task.c +2 -1
index e2d42593ce52..4cd2f9611c4a 100644
@@ -12,13 +12,14 @@ void sas_ssp_task_response(struct device *dev, struct sas_task *task,
 			   struct ssp_response_iu *iu)
 {
 	struct task_status_struct *tstat = &task->task_status;
+	u8 *resp = iu->resp_data;
 
 	tstat->resp = SAS_TASK_COMPLETE;
 
 	if (iu->datapres == 0)
 		tstat->stat = iu->status;
 	else if (iu->datapres == 1)
-		tstat->stat = iu->resp_data[3];
+		tstat->stat = resp[3];
 	else if (iu->datapres == 2) {
 		tstat->stat = SAM_STAT_CHECK_CONDITION;
 		tstat->buf_valid_size =
Signature · 1 line
2.26.0
Reviewed-by on v1 13/15
#33 lore raw

Re: [PATCH 13/15] scsi: sas: avoid gcc-10 zero-length-bounds warning

Quoted 21 lines · Arnd Bergmann

Two files access the zero-length resp_data[] array, which now causes a compiler warning:

drivers/scsi/aic94xx/aic94xx_tmf.c: In function 'asd_get_tmf_resp_tasklet':
drivers/scsi/aic94xx/aic94xx_tmf.c:291:22: warning: array subscript 3 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]

   291 |   res = ru->resp_data[3];
       |         ~~~~~~~~~~~~~^~~
In file included from include/scsi/libsas.h:15,
                  from drivers/scsi/aic94xx/aic94xx.h:16,
                  from drivers/scsi/aic94xx/aic94xx_tmf.c:11:
include/scsi/sas.h:557:9: note: while referencing 'resp_data'
   557 |  u8     resp_data[0];
       |         ^~~~~~~~~
drivers/scsi/libsas/sas_task.c: In function 'sas_ssp_task_response':
drivers/scsi/libsas/sas_task.c:21:30: warning: array subscript 3 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]
    21 |   tstat->stat = iu->resp_data[3];
       |                 ~~~~~~~~~~~~~^~~

In file included from include/scsi/scsi_transport_sas.h:8,
from drivers/scsi/libsas/sas_internal.h:14, from drivers/scsi/libsas/sas_task.c:3:
include/scsi/sas.h:557:9: note: while referencing 'resp_data'

Quoted 7 lines
   557 |  u8     resp_data[0];
       |         ^~~~~~~~~

This should really be a flexible-array member, but the structure already has such a member, swapping it out with sense_data[] would cause many more warnings elsewhere.

Hi Arnd,

If we really prefer flexible-array members over zero-length array members, then could we have a union of flexible-array members? I'm not sure if that's a good idea TBH (or even permitted), as these structures are defined by the SAS spec and good practice to keep as consistent as possible, but just wondering.

Apart from that:

Quoted 55 lines

As a workaround, add a temporary pointer that can be accessed without a warning.

Fixes 2908d778ab3e ("[SCSI] aic94xx: new driver")
Fixes 366ca51f30de ("[SCSI] libsas: abstract STP task status into a function")
---
  drivers/scsi/aic94xx/aic94xx_tmf.c | 4 +++-
  drivers/scsi/libsas/sas_task.c     | 3 ++-
  2 files changed, 5 insertions(+), 2 deletions(-)
drivers/scsi/aic94xx/aic94xx_tmf.c +3 -1
index f814026f26fa..a3139f9766c8 100644
@@ -269,6 +269,7 @@ static int asd_get_tmf_resp_tasklet(struct asd_ascb *ascb,
  	struct ssp_frame_hdr *fh;
  	struct ssp_response_iu   *ru;
  	int res = TMF_RESP_FUNC_FAILED;
+	u8 *resp;
 
  	ASD_DPRINTK("tmf resp tasklet\n");
 
@@ -287,8 +288,9 @@ static int asd_get_tmf_resp_tasklet(struct asd_ascb *ascb,
  	fh = edb->vaddr + 16;
  	ru = edb->vaddr + 16 + sizeof(*fh);
  	res = ru->status;
+	resp = ru->resp_data;
  	if (ru->datapres == 1)	  /* Response data present */
-		res = ru->resp_data[3];
+		res = resp[3];
  #if 0
  	ascb->tag = fh->tag;
  #endif
drivers/scsi/libsas/sas_task.c +2 -1
index e2d42593ce52..4cd2f9611c4a 100644
@@ -12,13 +12,14 @@ void sas_ssp_task_response(struct device *dev, struct sas_task *task,
  			   struct ssp_response_iu *iu)
  {
  	struct task_status_struct *tstat = &task->task_status;
+	u8 *resp = iu->resp_data;
 
  	tstat->resp = SAS_TASK_COMPLETE;
 
  	if (iu->datapres == 0)
  		tstat->stat = iu->status;
  	else if (iu->datapres == 1)
-		tstat->stat = iu->resp_data[3];
+		tstat->stat = resp[3];
  	else if (iu->datapres == 2) {
  		tstat->stat = SAM_STAT_CHECK_CONDITION;
  		tstat->buf_valid_size =
#34 lore raw

Re: [PATCH 13/15] scsi: sas: avoid gcc-10 zero-length-bounds warning

Quoted 8 lines · John Garry
Quoted 4 lines · Arnd Bergmann

This should really be a flexible-array member, but the structure already has such a member, swapping it out with sense_data[] would cause many more warnings elsewhere.

Quoted 7 lines

Hi Arnd,

If we really prefer flexible-array members over zero-length array members, then could we have a union of flexible-array members? I'm not sure if that's a good idea TBH (or even permitted), as these structures are defined by the SAS spec and good practice to keep as consistent as possible, but just wondering.

gcc does not allow flexible-array members inside of a union, or more than one flexible-array member at the end of a structure.

I found one hack that would work, but I think it's too ugly and likely not well-defined either:

struct ssp_response_iu {
...
struct {
u8 dummy[0]; /* a struct must have at least one
non-flexible member */

                u8      resp_data[]; /* allowed here because it's at
the one of a struct */
        };
        u8     sense_data[];
} __attribute__ ((packed));
Quoted 3 lines

Apart from that:

Thanks!

Arnd

#35 lore raw

Re: [PATCH 13/15] scsi: sas: avoid gcc-10 zero-length-bounds warning

Quoted 20 lines · Arnd Bergmann
Quoted 7 lines · John Garry
Quoted 4 lines · Arnd Bergmann

This should really be a flexible-array member, but the structure already has such a member, swapping it out with sense_data[] would cause many more warnings elsewhere.

Quoted 7 lines

Hi Arnd,

If we really prefer flexible-array members over zero-length array members, then could we have a union of flexible-array members? I'm not sure if that's a good idea TBH (or even permitted), as these structures are defined by the SAS spec and good practice to keep as consistent as possible, but just wondering.

gcc does not allow flexible-array members inside of a union, or more than one flexible-array member at the end of a structure.

Quoted 8 lines

I found one hack that would work, but I think it's too ugly and likely not well-defined either:

struct ssp_response_iu {
...
struct {
u8 dummy[0]; /* a struct must have at least one
non-flexible member */

If gcc is now warning about zero length members, why isn't it warning about this one ... are unions temporarily excluded?

Quoted 5 lines
                u8      resp_data[]; /* allowed here because it's at
the one of a struct */
        };
        u8     sense_data[];
} __attribute__ ((packed));

Let's go back to what the standard says: we want the data beyond the ssp_response_iu to be addressable either as sense_data if it's an error return or resp_data if it's a real response. What about trying to use an alias attribute inside the structure ... will that work on gcc-10?

James

#36 lore raw

Re: [PATCH 13/15] scsi: sas: avoid gcc-10 zero-length-bounds warning

Quoted 6 lines · James Bottomley
Quoted 5 lines · Arnd Bergmann

On Fri, May 1, 2020 at 9:48 AM John Garry <[email protected]> wrote:
I found one hack that would work, but I think it's too ugly and likely not well-defined either:

Quoted 8 lines
Quoted 5 lines

struct ssp_response_iu {
...
struct {
u8 dummy[0]; /* a struct must have at least one
non-flexible member */

If gcc is now warning about zero length members, why isn't it warning about this one ... are unions temporarily excluded?

It does not warn about all zero-length arrays, but it does warn when you try to access an array with an out-of-range index, and this apparently got extended in gcc-10 to any index for zero-length arrays.

Quoted 10 lines
Quoted 5 lines
                u8      resp_data[]; /* allowed here because it's at
the one of a struct */
        };
        u8     sense_data[];
} __attribute__ ((packed));

Let's go back to what the standard says: we want the data beyond the ssp_response_iu to be addressable either as sense_data if it's an error return or resp_data if it's a real response. What about trying to use an alias attribute inside the structure ... will that work on gcc-10?

I think alias attributes only work for functions and variables, but not for struct members.

A "#define sense_data resp_data" would obviously work, but it's rather error-prone when other code uses the same identifiers.

Another option would be an inline helper like

static inline u8 *ssp_response_data(struct ssp_response_iu *iu)

{
      return iu.resp_data;
}

Arnd

#37 lore raw

PATCH 14/15 isci: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

gcc-10 warns about accesses to zero-length arrays:

drivers/scsi/isci/task.h:125:31: error: array subscript 0 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Werror=zero-length-bounds]

  125 |    tmf->resp.resp_iu.resp_data[0],
      |    ~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
include/linux/dynamic_debug.h:143:2: note: in expansion of macro '__dynamic_func_call'
  143 |  __dynamic_func_call(__UNIQUE_ID(ddebug), fmt, func, ##__VA_ARGS__)
      |  ^~~~~~~~~~~~~~~~~~~
include/linux/dynamic_debug.h:157:2: note: in expansion of macro '_dynamic_func_call'
  157 |  _dynamic_func_call(fmt,__dynamic_dev_dbg,   \
      |  ^~~~~~~~~~~~~~~~~~
include/linux/dev_printk.h:115:2: note: in expansion of macro 'dynamic_dev_dbg'
  115 |  dynamic_dev_dbg(dev, dev_fmt(fmt), ##__VA_ARGS__)
      |  ^~~~~~~~~~~~~~~
drivers/scsi/isci/task.h:111:3: note: in expansion of macro 'dev_dbg'
  111 |   dev_dbg(&ihost->pdev->dev,
      |   ^~~~~~~
In file included from include/scsi/libsas.h:15,
                 from drivers/scsi/isci/task.c:59:
include/scsi/sas.h:326:9: note: while referencing 'resp_data'
  326 |  u8     resp_data[0];
      |         ^~~~~~~~~

This instance is not a bug, so just work around the warning by adding a temporary pointer.

---
 drivers/scsi/isci/task.h | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/scsi/isci/task.h +3 -4
index 8f4531f22ac2..2d556d7b5292 100644
@@ -98,6 +98,8 @@ struct isci_tmf {
 
 static inline void isci_print_tmf(struct isci_host *ihost, struct isci_tmf *tmf)
 {
+	u8 *resp = tmf->resp.resp_iu.resp_data;
+
 	if (SAS_PROTOCOL_SATA == tmf->proto)
 		dev_dbg(&ihost->pdev->dev,
 			"%s: status = %x\n"
@@ -122,10 +124,7 @@ static inline void isci_print_tmf(struct isci_host *ihost, struct isci_tmf *tmf)
 			tmf->resp.resp_iu.datapres,
 			tmf->resp.resp_iu.status,
 			be32_to_cpu(tmf->resp.resp_iu.response_data_len),
-			tmf->resp.resp_iu.resp_data[0],
-			tmf->resp.resp_iu.resp_data[1],
-			tmf->resp.resp_iu.resp_data[2],
-			tmf->resp.resp_iu.resp_data[3]);
+			resp[0], resp[1], resp[2], resp[3]);
 }
 
 
Signature · 1 line
2.26.0
#38 lore raw

PATCH 15/15 nvme: avoid gcc-10 zero-length-bounds warning

In reply to Arnd Bergmann

When CONFIG_ARCH_NO_SG_CHAIN is set, op->sgl[0] cannot be dereferenced, as gcc-10 now points out:

drivers/nvme/host/fc.c: In function 'nvme_fc_init_request':
drivers/nvme/host/fc.c:1774:29: warning: array subscript 0 is outside the bounds of an interior zero-length array 'struct scatterlist[0]' [-Wzero-length-bounds]

 1774 |  op->op.fcp_req.first_sgl = &op->sgl[0];
      |                             ^~~~~~~~~~~

drivers/nvme/host/fc.c:98:21: note: while referencing 'sgl'
98 | struct scatterlist sgl[NVME_INLINE_SG_CNT];
| ^~~

I don't know if this is a legitimate warning or a false-positive. If this is just a false alarm, the warning is easily suppressed by interpreting the array as a pointer.

Fixes b1ae1a238900 ("nvme-fc: Avoid preallocating big SGL for data")
---
 drivers/nvme/host/fc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/nvme/host/fc.c +1 -1
index 887537d1243f..b7ecda48d597 100644
@@ -1771,7 +1771,7 @@ nvme_fc_init_request(struct blk_mq_tag_set *set, struct request *rq,
 	res = __nvme_fc_init_request(ctrl, queue, &op->op, rq, queue->rqcnt++);
 	if (res)
 		return res;
-	op->op.fcp_req.first_sgl = &op->sgl[0];
+	op->op.fcp_req.first_sgl = op->sgl;
 	op->op.fcp_req.private = &op->priv[0];
 	nvme_req(rq)->ctrl = &ctrl->ctrl;
 	return res;
Signature · 7 lines
2.26.0


_______________________________________________
linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
#39 lore raw

Re: [PATCH 15/15] nvme: avoid gcc-10 zero-length-bounds warning

Quoted 7 lines · Arnd Bergmann

When CONFIG_ARCH_NO_SG_CHAIN is set, op->sgl[0] cannot be dereferenced, as gcc-10 now points out:

drivers/nvme/host/fc.c: In function 'nvme_fc_init_request':
drivers/nvme/host/fc.c:1774:29: warning: array subscript 0 is outside the bounds of an interior zero-length array 'struct scatterlist[0]' [-Wzero-length-bounds]

 1774 |  op->op.fcp_req.first_sgl = &op->sgl[0];
      |                             ^~~~~~~~~~~
Quoted 7 lines

drivers/nvme/host/fc.c:98:21: note: while referencing 'sgl'
98 | struct scatterlist sgl[NVME_INLINE_SG_CNT];
| ^~~

I don't know if this is a legitimate warning or a false-positive. If this is just a false alarm, the warning is easily suppressed by interpreting the array as a pointer.

This looks like a surpression to be, but then again I find the new code actually cleaner, so I'm fine with it :)