Hello,
First, my apologies for the long email.
I am requesting guidance on how to approach resolving the zero element flexible
VLO struct implementation in this driver in file drivers/staging/waln-ng/hfa384x.f
The struct hfa384x_pdrec contains nested structs with zero element arrays. These
zero element structs are part of a union 'data' inside the struct container. This
union 'data' is the last element of this container. Please see the code snip below:
<snip>
1068 struct hfa384x_pdrec {
1 __le16 len; /* in words */
2 __le16 code;
3 union pdr {
4 struct hfa384x_pdr_pcb_partnum pcb_partnum;
11 struct hfa384x_pdr_nicid nicid;
12 struct hfa384x_pdr_refdac_measurements refdac_measurements;
13 struct hfa384x_pdr_vgdac_measurements vgdac_measurements;
14 struct hfa384x_pdr_level_comp_measurements level_compc_measurements;
15 struct hfa384x_pdr_mac_address mac_address;
39 } data;
40 } __packed;
</snip>
The three structures in question are declared as follows in the same file:
<snip>
962 struct hfa384x_pdr_refdac_measurements {
1 u16 value[0];
2 } __packed;
3
4 struct hfa384x_pdr_vgdac_measurements {
5 u16 value[0];
6 } __packed;
7
8 struct hfa384x_pdr_level_comp_measurements {
9 u16 value[0];
10 } __packed;
</snip>
As per the C99 specifications, the flexible array struct should have at least
one member other than the true flexible array member. So converting these from
[0] to [] is not feasible in the current form.
I did not find these struct variables being used for memory allocation in the
code directly. My find may be short since the implementation appears to get very
complex as I tried to get deeper.
Can you please suggest how should I approach correcting the zero element flex
array implementation here? Can these structs be removed if they are unused?
Thank you.
./drv
On Tue, Nov 08, 2022 at 08:42:59PM +0530, Deepak R Varma wrote:
> Hello,
>
> First, my apologies for the long email.
> I am requesting guidance on how to approach resolving the zero element flexible
> VLO struct implementation in this driver in file drivers/staging/waln-ng/hfa384x.f
>
> The struct hfa384x_pdrec contains nested structs with zero element arrays. These
> zero element structs are part of a union 'data' inside the struct container. This
> union 'data' is the last element of this container. Please see the code snip below:
>
> <snip>
>
> 1068 struct hfa384x_pdrec {
> 1 __le16 len; /* in words */
> 2 __le16 code;
> 3 union pdr {
> 4 struct hfa384x_pdr_pcb_partnum pcb_partnum;
> 11 struct hfa384x_pdr_nicid nicid;
> 12 struct hfa384x_pdr_refdac_measurements refdac_measurements;
> 13 struct hfa384x_pdr_vgdac_measurements vgdac_measurements;
> 14 struct hfa384x_pdr_level_comp_measurements level_compc_measurements;
> 15 struct hfa384x_pdr_mac_address mac_address;
> 39 } data;
> 40 } __packed;
>
> </snip>
>
> The three structures in question are declared as follows in the same file:
>
> <snip>
> 962 struct hfa384x_pdr_refdac_measurements {
> 1 u16 value[0];
> 2 } __packed;
> 3
> 4 struct hfa384x_pdr_vgdac_measurements {
> 5 u16 value[0];
> 6 } __packed;
> 7
> 8 struct hfa384x_pdr_level_comp_measurements {
> 9 u16 value[0];
> 10 } __packed;
> </snip>
>
> As per the C99 specifications, the flexible array struct should have at least
> one member other than the true flexible array member. So converting these from
> [0] to [] is not feasible in the current form.
>
> I did not find these struct variables being used for memory allocation in the
> code directly. My find may be short since the implementation appears to get very
> complex as I tried to get deeper.
>
> Can you please suggest how should I approach correcting the zero element flex
> array implementation here? Can these structs be removed if they are unused?
Are you sure they are unused?
They look like structures that are read from the memory of a device,
right? Try removing the structures from the union and see what happens
:)
thanks,
greg k-h
On Tue, Nov 08, 2022 at 09:15:10PM +0530, Deepak R Varma wrote:
> On Tue, Nov 08, 2022 at 04:34:15PM +0100, Greg Kroah-Hartman wrote:
> > On Tue, Nov 08, 2022 at 08:42:59PM +0530, Deepak R Varma wrote:
> > > Hello,
> > >
> > > First, my apologies for the long email.
> > > I am requesting guidance on how to approach resolving the zero element flexible
> > > VLO struct implementation in this driver in file drivers/staging/waln-ng/hfa384x.f
> > >
> > > The struct hfa384x_pdrec contains nested structs with zero element arrays. These
> > > zero element structs are part of a union 'data' inside the struct container. This
> > > union 'data' is the last element of this container. Please see the code snip below:
> > >
> > > <snip>
> > >
> > > 1068 struct hfa384x_pdrec {
> > > 1 __le16 len; /* in words */
> > > 2 __le16 code;
> > > 3 union pdr {
> > > 4 struct hfa384x_pdr_pcb_partnum pcb_partnum;
> > > 11 struct hfa384x_pdr_nicid nicid;
> > > 12 struct hfa384x_pdr_refdac_measurements refdac_measurements;
> > > 13 struct hfa384x_pdr_vgdac_measurements vgdac_measurements;
> > > 14 struct hfa384x_pdr_level_comp_measurements level_compc_measurements;
> > > 15 struct hfa384x_pdr_mac_address mac_address;
> > > 39 } data;
> > > 40 } __packed;
> > >
> > > </snip>
> > >
> > > The three structures in question are declared as follows in the same file:
> > >
> > > <snip>
> > > 962 struct hfa384x_pdr_refdac_measurements {
> > > 1 u16 value[0];
> > > 2 } __packed;
> > > 3
> > > 4 struct hfa384x_pdr_vgdac_measurements {
> > > 5 u16 value[0];
> > > 6 } __packed;
> > > 7
> > > 8 struct hfa384x_pdr_level_comp_measurements {
> > > 9 u16 value[0];
> > > 10 } __packed;
> > > </snip>
> > >
> > > As per the C99 specifications, the flexible array struct should have at least
> > > one member other than the true flexible array member. So converting these from
> > > [0] to [] is not feasible in the current form.
> > >
> > > I did not find these struct variables being used for memory allocation in the
> > > code directly. My find may be short since the implementation appears to get very
> > > complex as I tried to get deeper.
> > >
> > > Can you please suggest how should I approach correcting the zero element flex
> > > array implementation here? Can these structs be removed if they are unused?
> >
> > Are you sure they are unused?
> >
> > They look like structures that are read from the memory of a device,
> > right? Try removing the structures from the union and see what happens
> > :)
>
> I did remove the structs from the union and it built fine. Is there anything else I
> can check/test to verify the impact?
>
> <snip>
> drv@qemulion:~/git/kernels/staging$ git diff
> diff --git a/drivers/staging/wlan-ng/hfa384x.h b/drivers/staging/wlan-ng/hfa384x.h
> index 0611e37df6ac..8fe10aa93dfb 100644
> --- a/drivers/staging/wlan-ng/hfa384x.h
> +++ b/drivers/staging/wlan-ng/hfa384x.h
> @@ -1077,9 +1077,6 @@ struct hfa384x_pdrec {
> struct hfa384x_pdr_mfisuprange mfisuprange;
> struct hfa384x_pdr_cfisuprange cfisuprange;
> struct hfa384x_pdr_nicid nicid;
> - struct hfa384x_pdr_refdac_measurements refdac_measurements;
> - struct hfa384x_pdr_vgdac_measurements vgdac_measurements;
> - struct hfa384x_pdr_level_comp_measurements level_compc_measurements;
> struct hfa384x_pdr_mac_address mac_address;
> struct hfa384x_pdr_mkk_callname mkk_callname;
> struct hfa384x_pdr_regdomain regdomain;
> drv@qemulion:~/git/kernels/staging$ make M=drivers/staging/wlan-ng/
> CC [M] drivers/staging/wlan-ng/prism2usb.o
> CC [M] drivers/staging/wlan-ng/p80211netdev.o
> LD [M] drivers/staging/wlan-ng/prism2_usb.o
> MODPOST drivers/staging/wlan-ng/Module.symvers
> LD [M] drivers/staging/wlan-ng/prism2_usb.ko
> BTF [M] drivers/staging/wlan-ng/prism2_usb.ko
> drv@qemulion:~/git/kernels/staging$
> </snip>
>
Test the device to make sure it still works?
On Tue, Nov 08, 2022 at 04:34:15PM +0100, Greg Kroah-Hartman wrote:
> On Tue, Nov 08, 2022 at 08:42:59PM +0530, Deepak R Varma wrote:
> > Hello,
> >
> > First, my apologies for the long email.
> > I am requesting guidance on how to approach resolving the zero element flexible
> > VLO struct implementation in this driver in file drivers/staging/waln-ng/hfa384x.f
> >
> > The struct hfa384x_pdrec contains nested structs with zero element arrays. These
> > zero element structs are part of a union 'data' inside the struct container. This
> > union 'data' is the last element of this container. Please see the code snip below:
> >
> > <snip>
> >
> > 1068 struct hfa384x_pdrec {
> > 1 __le16 len; /* in words */
> > 2 __le16 code;
> > 3 union pdr {
> > 4 struct hfa384x_pdr_pcb_partnum pcb_partnum;
> > 11 struct hfa384x_pdr_nicid nicid;
> > 12 struct hfa384x_pdr_refdac_measurements refdac_measurements;
> > 13 struct hfa384x_pdr_vgdac_measurements vgdac_measurements;
> > 14 struct hfa384x_pdr_level_comp_measurements level_compc_measurements;
> > 15 struct hfa384x_pdr_mac_address mac_address;
> > 39 } data;
> > 40 } __packed;
> >
> > </snip>
> >
> > The three structures in question are declared as follows in the same file:
> >
> > <snip>
> > 962 struct hfa384x_pdr_refdac_measurements {
> > 1 u16 value[0];
> > 2 } __packed;
> > 3
> > 4 struct hfa384x_pdr_vgdac_measurements {
> > 5 u16 value[0];
> > 6 } __packed;
> > 7
> > 8 struct hfa384x_pdr_level_comp_measurements {
> > 9 u16 value[0];
> > 10 } __packed;
> > </snip>
> >
> > As per the C99 specifications, the flexible array struct should have at least
> > one member other than the true flexible array member. So converting these from
> > [0] to [] is not feasible in the current form.
> >
> > I did not find these struct variables being used for memory allocation in the
> > code directly. My find may be short since the implementation appears to get very
> > complex as I tried to get deeper.
> >
> > Can you please suggest how should I approach correcting the zero element flex
> > array implementation here? Can these structs be removed if they are unused?
>
> Are you sure they are unused?
>
> They look like structures that are read from the memory of a device,
> right? Try removing the structures from the union and see what happens
> :)
I did remove the structs from the union and it built fine. Is there anything else I
can check/test to verify the impact?
<snip>
drv@qemulion:~/git/kernels/staging$ git diff
diff --git a/drivers/staging/wlan-ng/hfa384x.h b/drivers/staging/wlan-ng/hfa384x.h
index 0611e37df6ac..8fe10aa93dfb 100644
--- a/drivers/staging/wlan-ng/hfa384x.h
+++ b/drivers/staging/wlan-ng/hfa384x.h
@@ -1077,9 +1077,6 @@ struct hfa384x_pdrec {
struct hfa384x_pdr_mfisuprange mfisuprange;
struct hfa384x_pdr_cfisuprange cfisuprange;
struct hfa384x_pdr_nicid nicid;
- struct hfa384x_pdr_refdac_measurements refdac_measurements;
- struct hfa384x_pdr_vgdac_measurements vgdac_measurements;
- struct hfa384x_pdr_level_comp_measurements level_compc_measurements;
struct hfa384x_pdr_mac_address mac_address;
struct hfa384x_pdr_mkk_callname mkk_callname;
struct hfa384x_pdr_regdomain regdomain;
drv@qemulion:~/git/kernels/staging$ make M=drivers/staging/wlan-ng/
CC [M] drivers/staging/wlan-ng/prism2usb.o
CC [M] drivers/staging/wlan-ng/p80211netdev.o
LD [M] drivers/staging/wlan-ng/prism2_usb.o
MODPOST drivers/staging/wlan-ng/Module.symvers
LD [M] drivers/staging/wlan-ng/prism2_usb.ko
BTF [M] drivers/staging/wlan-ng/prism2_usb.ko
drv@qemulion:~/git/kernels/staging$
</snip>
>
> thanks,
>
> greg k-h
On Tue, Nov 08, 2022 at 04:52:44PM +0100, Greg Kroah-Hartman wrote:
> On Tue, Nov 08, 2022 at 09:15:10PM +0530, Deepak R Varma wrote:
> > On Tue, Nov 08, 2022 at 04:34:15PM +0100, Greg Kroah-Hartman wrote:
> > MODPOST drivers/staging/wlan-ng/Module.symvers
> > LD [M] drivers/staging/wlan-ng/prism2_usb.ko
> > BTF [M] drivers/staging/wlan-ng/prism2_usb.ko
> > drv@qemulion:~/git/kernels/staging$
> > </snip>
> >
>
> Test the device to make sure it still works?
I was able to build and load the driver on my machine. I do not have p54 device
to test. Is there another way to test it? Some sort of a udev program???
<snip>
static struct usb_driver prism2_usb_driver = {
- .name = "prism2_usb",
+ .name = "prism2_usb_dvk",
.probe = prism2sta_probe_usb,
.disconnect = prism2sta_disconnect_usb,
.id_table = usb_prism_tbl,
drv@qemulion:~/git/kernels/staging$ sudo dmesg
[ 948.476144] prism2_usb: module is from the staging directory, the quality is unknown, you have been warned.
[ 948.478631] usbcore: registered new interface driver prism2_usb_dvk
</snip>
Thank you,
./drv
>
On Tue, Nov 08, 2022 at 11:07:23PM +0530, Deepak R Varma wrote:
> On Tue, Nov 08, 2022 at 04:52:44PM +0100, Greg Kroah-Hartman wrote:
> > On Tue, Nov 08, 2022 at 09:15:10PM +0530, Deepak R Varma wrote:
> > > On Tue, Nov 08, 2022 at 04:34:15PM +0100, Greg Kroah-Hartman wrote:
> > > MODPOST drivers/staging/wlan-ng/Module.symvers
> > > LD [M] drivers/staging/wlan-ng/prism2_usb.ko
> > > BTF [M] drivers/staging/wlan-ng/prism2_usb.ko
> > > drv@qemulion:~/git/kernels/staging$
> > > </snip>
> > >
> >
> > Test the device to make sure it still works?
>
> I was able to build and load the driver on my machine. I do not have p54 device
> to test. Is there another way to test it? Some sort of a udev program???
You need the real hardware to test it properly.
On Tue, Nov 08, 2022 at 07:32:19PM +0100, Greg Kroah-Hartman wrote:
> On Tue, Nov 08, 2022 at 11:07:23PM +0530, Deepak R Varma wrote:
> > On Tue, Nov 08, 2022 at 04:52:44PM +0100, Greg Kroah-Hartman wrote:
> > > On Tue, Nov 08, 2022 at 09:15:10PM +0530, Deepak R Varma wrote:
> > > > On Tue, Nov 08, 2022 at 04:34:15PM +0100, Greg Kroah-Hartman wrote:
> > > > MODPOST drivers/staging/wlan-ng/Module.symvers
> > > > LD [M] drivers/staging/wlan-ng/prism2_usb.ko
> > > > BTF [M] drivers/staging/wlan-ng/prism2_usb.ko
> > > > drv@qemulion:~/git/kernels/staging$
> > > > </snip>
> > > >
> > >
> > > Test the device to make sure it still works?
> >
> > I was able to build and load the driver on my machine. I do not have p54 device
> > to test. Is there another way to test it? Some sort of a udev program???
>
> You need the real hardware to test it properly.
Do you know if I work with someone to test the change locally rather than
sending in a untested change?
Thank you,
./drv
>
On Wed, Nov 09, 2022 at 12:08:41AM +0530, Deepak R Varma wrote:
> On Tue, Nov 08, 2022 at 07:32:19PM +0100, Greg Kroah-Hartman wrote:
> > On Tue, Nov 08, 2022 at 11:07:23PM +0530, Deepak R Varma wrote:
> > > On Tue, Nov 08, 2022 at 04:52:44PM +0100, Greg Kroah-Hartman wrote:
> > > > On Tue, Nov 08, 2022 at 09:15:10PM +0530, Deepak R Varma wrote:
> > > > > On Tue, Nov 08, 2022 at 04:34:15PM +0100, Greg Kroah-Hartman wrote:
> > > > > MODPOST drivers/staging/wlan-ng/Module.symvers
> > > > > LD [M] drivers/staging/wlan-ng/prism2_usb.ko
> > > > > BTF [M] drivers/staging/wlan-ng/prism2_usb.ko
> > > > > drv@qemulion:~/git/kernels/staging$
> > > > > </snip>
> > > > >
> > > >
> > > > Test the device to make sure it still works?
> > >
> > > I was able to build and load the driver on my machine. I do not have p54 device
> > > to test. Is there another way to test it? Some sort of a udev program???
> >
> > You need the real hardware to test it properly.
>
> Do you know if I work with someone to test the change locally rather than
> sending in a untested change?
If you can find someone, sure!
On Tue, Nov 08, 2022 at 08:43:52PM +0100, Greg Kroah-Hartman wrote:
> On Wed, Nov 09, 2022 at 12:08:41AM +0530, Deepak R Varma wrote:
> > On Tue, Nov 08, 2022 at 07:32:19PM +0100, Greg Kroah-Hartman wrote:
> > > > > Test the device to make sure it still works?
> > > >
> > > > I was able to build and load the driver on my machine. I do not have p54 device
> > > > to test. Is there another way to test it? Some sort of a udev program???
> > >
> > > You need the real hardware to test it properly.
> >
> > Do you know if I work with someone to test the change locally rather than
> > sending in a untested change?
>
> If you can find someone, sure!
Hello Greg,
I reached out to couple of recent patch authors to check if they are able to
help test the change.
Hello all,
Does anyone have a hardware that can test prism2_usb [wlan-ng] driver change? I
assume it would be quick to test the change I am proposing. Let me know if you
are able to help.
Thank you!
./drv
>
On Tue, Nov 08, 2022 at 08:42:59PM +0530, Deepak R Varma wrote:
>
> Can you please suggest how should I approach correcting the zero element flex
> array implementation here? Can these structs be removed if they are unused?
You can try using DECLARE_FLEX_ARRAY(). See this[1] patch.
--
Gustavo
[1] https://git.kernel.org/linus/6e4a53ee7989c8a2b9fc3b14cd90f6e2d613ca76
On Sun, Nov 13, 2022 at 12:33:43AM -0600, Gustavo A. R. Silva wrote:
> On Tue, Nov 08, 2022 at 08:42:59PM +0530, Deepak R Varma wrote:
> >
> > Can you please suggest how should I approach correcting the zero element flex
> > array implementation here? Can these structs be removed if they are unused?
>
> You can try using DECLARE_FLEX_ARRAY(). See this[1] patch.
Thank you very much for the quick reply. The link you provided is very helpful.
I will review it in detail and correct my patch proposal accordingly.
ps: Also. thank you for including the cc list. I has already made a lot of noise
with on the lists, so did not copy them intentionally in my email to you.
./drv
>
> --
> Gustavo
>
> [1] https://git.kernel.org/linus/6e4a53ee7989c8a2b9fc3b14cd90f6e2d613ca76
>