drivers/net/wireless/marvell/mwifiex/fw.h | 2 +- drivers/net/wireless/marvell/mwifiex/scan.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Replace one-element array with a flexible-array member in `struct
mwifiex_ie_types_wildcard_ssid_params` to fix the following warning
on a MT8173 Chromebook (mt8173-elm-hana):
[ 356.775250] ------------[ cut here ]------------
[ 356.784543] memcpy: detected field-spanning write (size 6) of single field "wildcard_ssid_tlv->ssid" at drivers/net/wireless/marvell/mwifiex/scan.c:904 (size 1)
[ 356.813403] WARNING: CPU: 3 PID: 742 at drivers/net/wireless/marvell/mwifiex/scan.c:904 mwifiex_scan_networks+0x4fc/0xf28 [mwifiex]
The "(size 6)" above is exactly the length of the SSID of the network
this device was connected to. The source of the warning looks like:
ssid_len = user_scan_in->ssid_list[i].ssid_len;
[...]
memcpy(wildcard_ssid_tlv->ssid,
user_scan_in->ssid_list[i].ssid, ssid_len);
Also adjust a #define that uses sizeof() on this struct to keep the
value same as before.
Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com>
---
I found these relevant patches that modify other such arrays, where the
second one removes a -1 from some sizeof() calculation:
https://lore.kernel.org/lkml/Y9xkECG3uTZ6T1dN@work/T/#u
https://lore.kernel.org/lkml/ZsZa5xRcsLq9D+RX@elsanto/T/#u
So I think we need the +1 to keep things same. But it appears to work
fine without it, so I'm not sure. Maybe it should've had a -1 before
that I would remove with this?
drivers/net/wireless/marvell/mwifiex/fw.h | 2 +-
drivers/net/wireless/marvell/mwifiex/scan.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/wireless/marvell/mwifiex/fw.h b/drivers/net/wireless/marvell/mwifiex/fw.h
index d03129d5d24e..4a96281792cc 100644
--- a/drivers/net/wireless/marvell/mwifiex/fw.h
+++ b/drivers/net/wireless/marvell/mwifiex/fw.h
@@ -875,7 +875,7 @@ struct mwifiex_ietypes_chanstats {
struct mwifiex_ie_types_wildcard_ssid_params {
struct mwifiex_ie_types_header header;
u8 max_ssid_length;
- u8 ssid[1];
+ u8 ssid[];
} __packed;
#define TSF_DATA_SIZE 8
diff --git a/drivers/net/wireless/marvell/mwifiex/scan.c b/drivers/net/wireless/marvell/mwifiex/scan.c
index cab889af4c4a..50af78ee935b 100644
--- a/drivers/net/wireless/marvell/mwifiex/scan.c
+++ b/drivers/net/wireless/marvell/mwifiex/scan.c
@@ -32,7 +32,7 @@
#define WILDCARD_SSID_TLV_MAX_SIZE \
(MWIFIEX_MAX_SSID_LIST_LENGTH * \
(sizeof(struct mwifiex_ie_types_wildcard_ssid_params) \
- + IEEE80211_MAX_SSID_LEN))
+ + IEEE80211_MAX_SSID_LEN + 1))
/* Maximum memory needed for a mwifiex_scan_cmd_config with all TLVs at max */
#define MAX_SCAN_CFG_ALLOC (sizeof(struct mwifiex_scan_cmd_config) \
base-commit: 4f3e012d4cfd1d9bf837870c961f462ca9f23ebe
--
2.45.2
On Tue, Sep 17, 2024 at 06:08:41PM +0300, Alper Nebi Yasak wrote: > Replace one-element array with a flexible-array member in `struct > mwifiex_ie_types_wildcard_ssid_params` to fix the following warning > on a MT8173 Chromebook (mt8173-elm-hana): > > [ 356.775250] ------------[ cut here ]------------ > [ 356.784543] memcpy: detected field-spanning write (size 6) of single field "wildcard_ssid_tlv->ssid" at drivers/net/wireless/marvell/mwifiex/scan.c:904 (size 1) > [ 356.813403] WARNING: CPU: 3 PID: 742 at drivers/net/wireless/marvell/mwifiex/scan.c:904 mwifiex_scan_networks+0x4fc/0xf28 [mwifiex] > > The "(size 6)" above is exactly the length of the SSID of the network > this device was connected to. The source of the warning looks like: > > ssid_len = user_scan_in->ssid_list[i].ssid_len; > [...] > memcpy(wildcard_ssid_tlv->ssid, > user_scan_in->ssid_list[i].ssid, ssid_len); > > Also adjust a #define that uses sizeof() on this struct to keep the > value same as before. > > Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com> > --- > I found these relevant patches that modify other such arrays, where the > second one removes a -1 from some sizeof() calculation: > > https://lore.kernel.org/lkml/Y9xkECG3uTZ6T1dN@work/T/#u > https://lore.kernel.org/lkml/ZsZa5xRcsLq9D+RX@elsanto/T/#u > > So I think we need the +1 to keep things same. But it appears to work > fine without it, so I'm not sure. Maybe it should've had a -1 before > that I would remove with this? Thanks for the investigation and patch! I believe I agree with the other comments, that then "+ 1" isn't necessary. It's just a wasteful extra byte of allocation. Can you send a v2? (Bonus: with the suggested Fixes tag. Double bonus if you test KASAN with __counted_by, for a second patch.) Thanks, Brian
On 9/17/2024 8:08 AM, Alper Nebi Yasak wrote: > Replace one-element array with a flexible-array member in `struct > mwifiex_ie_types_wildcard_ssid_params` to fix the following warning > on a MT8173 Chromebook (mt8173-elm-hana): > > [ 356.775250] ------------[ cut here ]------------ > [ 356.784543] memcpy: detected field-spanning write (size 6) of single field "wildcard_ssid_tlv->ssid" at drivers/net/wireless/marvell/mwifiex/scan.c:904 (size 1) > [ 356.813403] WARNING: CPU: 3 PID: 742 at drivers/net/wireless/marvell/mwifiex/scan.c:904 mwifiex_scan_networks+0x4fc/0xf28 [mwifiex] > > The "(size 6)" above is exactly the length of the SSID of the network > this device was connected to. The source of the warning looks like: > > ssid_len = user_scan_in->ssid_list[i].ssid_len; > [...] > memcpy(wildcard_ssid_tlv->ssid, > user_scan_in->ssid_list[i].ssid, ssid_len); > > Also adjust a #define that uses sizeof() on this struct to keep the > value same as before. > > Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com> > --- > I found these relevant patches that modify other such arrays, where the > second one removes a -1 from some sizeof() calculation: > > https://lore.kernel.org/lkml/Y9xkECG3uTZ6T1dN@work/T/#u > https://lore.kernel.org/lkml/ZsZa5xRcsLq9D+RX@elsanto/T/#u > > So I think we need the +1 to keep things same. But it appears to work > fine without it, so I'm not sure. Maybe it should've had a -1 before > that I would remove with this? I think the original code was incorrect and was allocating too much memory. I do not think WILDCARD_SSID_TLV_MAX_SIZE requires modification. > > drivers/net/wireless/marvell/mwifiex/fw.h | 2 +- > drivers/net/wireless/marvell/mwifiex/scan.c | 2 +- > 2 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/wireless/marvell/mwifiex/fw.h b/drivers/net/wireless/marvell/mwifiex/fw.h > index d03129d5d24e..4a96281792cc 100644 > --- a/drivers/net/wireless/marvell/mwifiex/fw.h > +++ b/drivers/net/wireless/marvell/mwifiex/fw.h > @@ -875,7 +875,7 @@ struct mwifiex_ietypes_chanstats { > struct mwifiex_ie_types_wildcard_ssid_params { > struct mwifiex_ie_types_header header; > u8 max_ssid_length; > - u8 ssid[1]; > + u8 ssid[]; > } __packed; > > #define TSF_DATA_SIZE 8 > diff --git a/drivers/net/wireless/marvell/mwifiex/scan.c b/drivers/net/wireless/marvell/mwifiex/scan.c > index cab889af4c4a..50af78ee935b 100644 > --- a/drivers/net/wireless/marvell/mwifiex/scan.c > +++ b/drivers/net/wireless/marvell/mwifiex/scan.c > @@ -32,7 +32,7 @@ > #define WILDCARD_SSID_TLV_MAX_SIZE \ > (MWIFIEX_MAX_SSID_LIST_LENGTH * \ > (sizeof(struct mwifiex_ie_types_wildcard_ssid_params) \ > - + IEEE80211_MAX_SSID_LEN)) > + + IEEE80211_MAX_SSID_LEN + 1)) > > /* Maximum memory needed for a mwifiex_scan_cmd_config with all TLVs at max */ > #define MAX_SCAN_CFG_ALLOC (sizeof(struct mwifiex_scan_cmd_config) \ > > base-commit: 4f3e012d4cfd1d9bf837870c961f462ca9f23ebe
On 17/09/24 20:10, Jeff Johnson wrote: > On 9/17/2024 8:08 AM, Alper Nebi Yasak wrote: >> Replace one-element array with a flexible-array member in `struct >> mwifiex_ie_types_wildcard_ssid_params` to fix the following warning >> on a MT8173 Chromebook (mt8173-elm-hana): >> >> [ 356.775250] ------------[ cut here ]------------ >> [ 356.784543] memcpy: detected field-spanning write (size 6) of single field "wildcard_ssid_tlv->ssid" at drivers/net/wireless/marvell/mwifiex/scan.c:904 (size 1) >> [ 356.813403] WARNING: CPU: 3 PID: 742 at drivers/net/wireless/marvell/mwifiex/scan.c:904 mwifiex_scan_networks+0x4fc/0xf28 [mwifiex] >> >> The "(size 6)" above is exactly the length of the SSID of the network >> this device was connected to. The source of the warning looks like: >> >> ssid_len = user_scan_in->ssid_list[i].ssid_len; >> [...] >> memcpy(wildcard_ssid_tlv->ssid, >> user_scan_in->ssid_list[i].ssid, ssid_len); >> >> Also adjust a #define that uses sizeof() on this struct to keep the >> value same as before. >> >> Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com> >> --- >> I found these relevant patches that modify other such arrays, where the >> second one removes a -1 from some sizeof() calculation: >> >> https://lore.kernel.org/lkml/Y9xkECG3uTZ6T1dN@work/T/#u >> https://lore.kernel.org/lkml/ZsZa5xRcsLq9D+RX@elsanto/T/#u >> >> So I think we need the +1 to keep things same. But it appears to work >> fine without it, so I'm not sure. Maybe it should've had a -1 before >> that I would remove with this? > > I think the original code was incorrect and was allocating too much memory. > I do not think WILDCARD_SSID_TLV_MAX_SIZE requires modification. If this is case, I think the changelog text should include the following 'Fixes' tag (and the `__counted_by` annotation added in a separate patch): Fixes: 5e6e3a92b9a4 ("wireless: mwifiex: initial commit for Marvell mwifiex driver") Thanks -- Gustavo > >> >> drivers/net/wireless/marvell/mwifiex/fw.h | 2 +- >> drivers/net/wireless/marvell/mwifiex/scan.c | 2 +- >> 2 files changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/net/wireless/marvell/mwifiex/fw.h b/drivers/net/wireless/marvell/mwifiex/fw.h >> index d03129d5d24e..4a96281792cc 100644 >> --- a/drivers/net/wireless/marvell/mwifiex/fw.h >> +++ b/drivers/net/wireless/marvell/mwifiex/fw.h >> @@ -875,7 +875,7 @@ struct mwifiex_ietypes_chanstats { >> struct mwifiex_ie_types_wildcard_ssid_params { >> struct mwifiex_ie_types_header header; >> u8 max_ssid_length; >> - u8 ssid[1]; >> + u8 ssid[]; >> } __packed; >> >> #define TSF_DATA_SIZE 8 >> diff --git a/drivers/net/wireless/marvell/mwifiex/scan.c b/drivers/net/wireless/marvell/mwifiex/scan.c >> index cab889af4c4a..50af78ee935b 100644 >> --- a/drivers/net/wireless/marvell/mwifiex/scan.c >> +++ b/drivers/net/wireless/marvell/mwifiex/scan.c >> @@ -32,7 +32,7 @@ >> #define WILDCARD_SSID_TLV_MAX_SIZE \ >> (MWIFIEX_MAX_SSID_LIST_LENGTH * \ >> (sizeof(struct mwifiex_ie_types_wildcard_ssid_params) \ >> - + IEEE80211_MAX_SSID_LEN)) >> + + IEEE80211_MAX_SSID_LEN + 1)) >> >> /* Maximum memory needed for a mwifiex_scan_cmd_config with all TLVs at max */ >> #define MAX_SCAN_CFG_ALLOC (sizeof(struct mwifiex_scan_cmd_config) \ >> >> base-commit: 4f3e012d4cfd1d9bf837870c961f462ca9f23ebe >
On Tue, Sep 17, 2024 at 06:08:41PM +0300, Alper Nebi Yasak wrote: > Replace one-element array with a flexible-array member in `struct > mwifiex_ie_types_wildcard_ssid_params` to fix the following warning > on a MT8173 Chromebook (mt8173-elm-hana): > > [ 356.775250] ------------[ cut here ]------------ > [ 356.784543] memcpy: detected field-spanning write (size 6) of single field "wildcard_ssid_tlv->ssid" at drivers/net/wireless/marvell/mwifiex/scan.c:904 (size 1) > [ 356.813403] WARNING: CPU: 3 PID: 742 at drivers/net/wireless/marvell/mwifiex/scan.c:904 mwifiex_scan_networks+0x4fc/0xf28 [mwifiex] > > The "(size 6)" above is exactly the length of the SSID of the network > this device was connected to. The source of the warning looks like: > > ssid_len = user_scan_in->ssid_list[i].ssid_len; > [...] > memcpy(wildcard_ssid_tlv->ssid, > user_scan_in->ssid_list[i].ssid, ssid_len); > > Also adjust a #define that uses sizeof() on this struct to keep the > value same as before. ... > #define WILDCARD_SSID_TLV_MAX_SIZE \ > (MWIFIEX_MAX_SSID_LIST_LENGTH * \ > (sizeof(struct mwifiex_ie_types_wildcard_ssid_params) \ > - + IEEE80211_MAX_SSID_LEN)) > + + IEEE80211_MAX_SSID_LEN + 1)) > This hunk has to be carefully checked by wireless people. I'm not sure that we need + 1 here. Otherwise, LGTM. -- With Best Regards, Andy Shevchenko
On 9/17/24 7:30 PM, Andy Shevchenko wrote: > This hunk has to be carefully checked by wireless people. I'm not sure > that we need + 1 here. Well, +1 looks redundant here indeed. And if you touch this anyway, please also consider '__counted_by(max_ssid_length)' for 'ssid' (and test your kernel with CONFIG_KASAN enabled of course). Dmitry
© 2016 - 2024 Red Hat, Inc.