Making pxe working in the "NAT" mode
PATCH 1/1 Making pxe working in the "NAT" mode
Hey folks,
I saw the lastest qemu support pxe so I tried it using the following command:
qemu -hda /tmp/a.img -net nic n- user -boot n
Note that now qemu support pxe mode, the mandatory -hda option is no more mandatory but this is not directly linked to my patch. I saw qemu integrates a dhcp & a tftp server but they don't seems to be configured/developped for a pxe boot process. Qemu says "no filename" and exits.
A friend of I asked me "how do you boot using pxe in qemu when you don't have a dhcp & a tftp server ?".
So I started to look how we can do that, I modified bootp to catch the field 60 of a dhcp request to find if its a pxe request or not. If yes, I set the filename to "/pxelinux.0", currently this is harcoded, maybe you can help me to find a better way to specify a bootloader.
Then, I've patched the tftp server to change the default rootdir in my home dir. "~/tftpboot". This is just for a Proof-of-concept, other methods could be fine. We can also imagine searching in a path restricted to the admin and then in the user dir. So it can give the admin the ability of defining some pxe boot for all the vm of a system or in the user dir if the user wants to have a special pxe boot scheme.
The last patch I did on the tftp server, is to allow the tsize negociation because pxelinux needs it.
This patch apply to 0.9.0, once it is applied and compiled here come the test procedure:
create a ~/tftpboot/pxelinux.cfg directory
copy pxelinux.0 in ~/tfptboot/
create a pxelinux configuration file called "default" in ~/tftpboot/pxelinux.cfg
Et voila ;)
Just call "qemu -hda /tmp/a.img -net nic n- user -boot n", it will start pxelinux.
I think this patch is very usefull because it give users & admin the ability to use pxe inside the VM without installing the full configuration (dhcp,tftp). This patch is mainly a proof-of-concept, this patch can be improved, comments and feedback are welcome.
bootp.c +14 -2
@@ -82,7 +82,7 @@ }static void dhcp_decode (const uint8_t *buf,int size, -int *pmsg_type) +int *pmsg_type,int *pxe_request) {const uint8_t *p, *p_end;int len, tag;
@@ -114,6 +114,11 @@if (len >=1 ) *pmsg_type = p[0 ];break ; +case RFC1533_CLASSIDENTIFIER : +if ((strncmp (p,"PXEClient" ,9 ) ==0 ) || (strncmp (p,"Etherboot" ,9 ) ==0 )) { + *pxe_request=1 ; + } +break ;default :break ; }
@@ -131,9 +136,10 @@struct in_addr dns_addr;int dhcp_msg_type, val;uint8_t *q; +int pxe_request=0 ;/* extract exact DHCP msg type */ -dhcp_decode (bp->bp_vend,DHCP_OPT_LEN , &dhcp_msg_type); +dhcp_decode (bp->bp_vend,DHCP_OPT_LEN , &dhcp_msg_type,&pxe_request);dprintf ("bootp packet op=%d msgtype=%d\n" , bp->bp_op, dhcp_msg_type);if (dhcp_msg_type ==0 )
@@ -168,6 +174,12 @@ goto new_addr; } } + +if (pxe_request==1 ) { +printf ("PXE Request Detected, setting filename\n" ); +strcpy (rbp->bp_file,"/pxelinux.0" ); + } +dprintf ("offered addr=%08x\n" ,ntohl (daddr.sin_addr.s_addr)); saddr.sin_addr.s_addr =htonl (ntohl (special_addr.s_addr) |CTL_ALIAS );
bootp.h +1 -0
@@ -57,6 +57,7 @@#define RFC1533_NBSCOPE 47#define RFC1533_XFS 48#define RFC1533_XDM 49 +#define RFC1533_CLASSIDENTIFIER 60#define RFC2132_REQ_ADDR 50#define RFC2132_LEASE_TIME 51
Seulement dans slirp: bootp.h~
Seulement dans slirp: debug.c~
tftp.c +80 -1
112 lines not on this page · raw message
tftp.h +1 -0
@@ -9,6 +9,7 @@#define TFTP_DATA 3#define TFTP_ACK 4#define TFTP_ERROR 5 +#define TFTP_OACK 6#define TFTP_FILENAME_MAX 512
Quoted 8 lines · Erwan Velu
Hey folks,
I saw the lastest qemu support pxe so I tried it using the following command:qemu -hda
/tmp/a.img-net nic n- user -boot nNote that now qemu support pxe mode, the mandatory -hda option is no more mandatory but this is not directly linked to my patch. I saw qemu integrates a dhcp & a tftp server but they don't seems to be configured/developped for a pxe boot process. Qemu says "no filename" and exits.
Hi Erwan,
I really like the idea of your patch. Of course, the hard coding is not really reasonable :-) Also, PXE is an x86-ism. What you've really done is added a mechanism to publish a BOOTP name. We do need to add a new option too as one may want to use -boot n without publishing a BOOTP filename.
Using your patch as a basis, I've written three patches. The first one is the tsize negotiation fix for the TFTP server. The second adds a -bootp option for specifying the location of the BOOTP image. The third changes the -tftp option to take root directory. The end result, is that now you can say:
qemu -hda /tmp/a.img -boot n -tftp ~/tftpboot -bootp /pxelinux.0
And you get the same results without the hard coding. What do you think?
Regards,
Anthony Liguori
Quoted 16 lines
A friend of I asked me "how do you boot using pxe in qemu when you don't have a dhcp & a tftp server ?".
So I started to look how we can do that, I modified bootp to catch the field 60 of a dhcp request to find if its a pxe request or not. If yes, I set the filename to "/pxelinux.0", currently this is harcoded, maybe you can help me to find a better way to specify a bootloader.Then, I've patched the tftp server to change the default rootdir in my home dir. "~/tftpboot". This is just for a Proof-of-concept, other methods could be fine. We can also imagine searching in a path restricted to the admin and then in the user dir. So it can give the admin the ability of defining some pxe boot for all the vm of a system or in the user dir if the user wants to have a special pxe boot scheme.
The last patch I did on the tftp server, is to allow the tsize negociation because pxelinux needs it.
This patch apply to 0.9.0, once it is applied and compiled here come the test procedure:
create a ~/tftpboot/pxelinux.cfg directory
copy pxelinux.0 in ~/tfptboot/
create a pxelinux configuration file called "default" in ~/tftpboot/pxelinux.cfgEt voila ;)
Quoted 9 lines
Just call "qemu -hda
/tmp/a.img-net nic n- user -boot n", it will start pxelinux.I think this patch is very usefull because it give users & admin the ability to use pxe inside the VM without installing the full configuration (dhcp,tftp). This patch is mainly a proof-of-concept, this patch can be improved, comments and feedback are welcome.
_______________________________________________
Qemu-devel mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/qemu-devel
slirp/bootp.c +6 -0
@@ -38,6 +38,8 @@ typedef struct {BOOTPClient bootp_clients[NB_ADDR ]; +const char *bootp_filename; +static const uint8_t rfc1533_cookie[] = {RFC1533_COOKIE };#ifdef DEBUG
@@ -168,6 +170,10 @@ static void bootp_reply(struct bootp_t * goto new_addr; } } + +if (bootp_filename) +snprintf (rbp->bp_file,sizeof (rbp->bp_file),"%s" , bootp_filename); +dprintf ("offered addr=%08x\n" ,ntohl (daddr.sin_addr.s_addr)); saddr.sin_addr.s_addr =htonl (ntohl (special_addr.s_addr) |CTL_ALIAS );
vl.c +6 -0
@@ -6488,6 +6488,7 @@ void help(void) "\n"#ifdef CONFIG_SLIRP "-tftp prefix allow tftp access to files starting with prefix [-net user]\n" +"-bootp file advertise file in BOOTP replies\n" #ifndef _WIN32"-smb dir allow SMB access to files in 'dir' [-net user]\n" #endif
@@ -6576,6 +6577,7 @@ enum { QEMU_OPTION_net, QEMU_OPTION_tftp, + QEMU_OPTION_bootp, QEMU_OPTION_smb, QEMU_OPTION_redir,
@@ -6652,6 +6654,7 @@ const QEMUOption qemu_options[] = { { "net", HAS_ARG, QEMU_OPTION_net}, #ifdef CONFIG_SLIRP { "tftp", HAS_ARG, QEMU_OPTION_tftp }, + { "bootp", HAS_ARG, QEMU_OPTION_bootp }, #ifndef _WIN32 { "smb", HAS_ARG, QEMU_OPTION_smb }, #endif
@@ -7185,6 +7188,9 @@ int main(int argc, char **argv)case QEMU_OPTION_tftp: tftp_prefix = optarg;break ; +case QEMU_OPTION_bootp: + bootp_filename = optarg; +break ;#ifndef _WIN32case QEMU_OPTION_smb:net_slirp_smb (optarg);
vl.h +1 -0
@@ -163,6 +163,7 @@ extern int no_quit;extern int no_quit;extern int semihosting_enabled;extern int autostart; +extern const char *bootp_filename;#define MAX_OPTION_ROMS 16extern const char *option_rom[MAX_OPTION_ROMS ];
qemu-doc.texi +10 -0
@@ -445,6 +445,16 @@ guest must be configured in binary mode guest must be configured in binary mode (use the command @code{bin} of the Unix TFTP client). The host IP address on the guest is as usual 10.0.2.2. + +@item -bootp file +When using the user mode network stack, broadcast @var{file} as the BOOTP +filename. In conjunction with @option{-tftp}, this can be used to network boot +a guest from a local directory. + +Example (using pxelinux): +@example +qemu -hda linux.img -boot n -tftp /path/to/tftp/files -bootp /pxelinux.0 +@end example @item -smb dir When using the user mode network stack, activate a built-in SMB
slirp/tftp.c +83 -0
101 lines not on this page · raw message
slirp/tftp.h +1 -0
@@ -9,6 +9,7 @@#define TFTP_DATA 3#define TFTP_ACK 4#define TFTP_ERROR 5 +#define TFTP_OACK 6#define TFTP_FILENAME_MAX 512
qemu-doc.texi +5 -6
@@ -438,13 +438,12 @@ override the default configuration (@opt override the default configuration (@option{-net nic -net user}) which is activated if no @option{-net} options are provided. -@item -tftp prefix +@item -tftp dir When using the user mode network stack, activate a built-in TFTP -server. All filenames beginning with @var{prefix} can be downloaded -from the host to the guest using a TFTP client. The TFTP client on the -guest must be configured in binary mode (use the command @code{bin} of -the Unix TFTP client). The host IP address on the guest is as usual -10.0.2.2. +server. The files in @var{dir} will be exposed as the root of a TFTP server. +The TFTP client on the guest must be configured in binary mode (use the command +@code{bin} of the Unix TFTP client). The host IP address on the guest is as +usual 10.0.2.2. @item -bootp file When using the user mode network stack, broadcast @var{file} as the BOOTP
slirp/tftp.c +10 -4
@@ -102,8 +102,15 @@ static int tftp_read_data(struct tftp_se {int fd;int bytes_read =0 ; - - fd =open (spt->filename,O_RDONLY |O_BINARY ); +char buffer[1024 ]; +int n; + + n =snprintf (buffer,sizeof (buffer),"%s/%s" , + tftp_prefix, spt->filename); +if (n >=sizeof (buffer)) +return -1 ; + + fd =open (buffer,O_RDONLY |O_BINARY );if (fd <0 ) {return -1 ;
@@ -328,8 +335,7 @@ static void tftp_handle_rrq(struct tftp_/* only allow exported prefixes */ -if (!tftp_prefix - || (strncmp (spt->filename, tftp_prefix,strlen (tftp_prefix)) !=0 )) { +if (!tftp_prefix) {tftp_send_error (spt,2 ,"Access violation" , tp);return ; }
vl.c +1 -1
@@ -6487,7 +6487,7 @@ void help(void)" is provided, the default is '-net nic -net user'\n" "\n" #ifdef CONFIG_SLIRP -"-tftp prefix allow tftp access to files starting with prefix [-net user]\n" +"-tftp dir allow tftp access to files in dir [-net user]\n" "-bootp file advertise file in BOOTP replies\n" #ifndef _WIN32"-smb dir allow SMB access to files in 'dir' [-net user]\n"
Quoted 9 lines · Anthony Liguori
Hi Erwan,
I really like the idea of your patch. Of course, the hard coding is not really reasonable :-) Also, PXE is an x86-ism. What you've really done is added a mechanism to publish a BOOTP name. We do need to add a new option too as one may want to use -boot n without publishing a BOOTP filename.
Using your patch as a basis, I've written three patches. The first one is the tsize negotiation fix for the TFTP server. The second adds a -bootp option for specifying the location of the BOOTP image. The third changes the -tftp option to take root directory. The end result, is that now you can say:
qemu -hda
/tmp/a.img-boot n -tftp ~/tftpboot -bootp /pxelinux.0And you get the same results without the hard coding. What do you think?
I just love it, you exactly did the stuff I didn't knew how to do. It just push a turn a simple Proof-of-concept into a "ready-to-use" feature.
Thanks a lot for you excellent work, I hope now this patch could be integrated upstream for allowing people to use PXE in qemu without the need of knowing how to configure the dhcp/tftp and bridging stuff.
One more time, thanks anthony for this patches.